2 * Copyright (C) 2005 - 2014 Emulex
5 * This program is free software; you can redistribute it and/or
6 * modify it under the terms of the GNU General Public License version 2
7 * as published by the Free Software Foundation. The full GNU General
8 * Public License is included in this distribution in the file called COPYING.
10 * Contact Information:
11 * linux-drivers@emulex.com
15 * Costa Mesa, CA 92626
18 #include <linux/prefetch.h>
19 #include <linux/module.h>
22 #include <asm/div64.h>
23 #include <linux/aer.h>
24 #include <linux/if_bridge.h>
25 #include <net/busy_poll.h>
26 #include <net/vxlan.h>
28 MODULE_VERSION(DRV_VER);
29 MODULE_DEVICE_TABLE(pci, be_dev_ids);
30 MODULE_DESCRIPTION(DRV_DESC " " DRV_VER);
31 MODULE_AUTHOR("Emulex Corporation");
32 MODULE_LICENSE("GPL");
34 static unsigned int num_vfs;
35 module_param(num_vfs, uint, S_IRUGO);
36 MODULE_PARM_DESC(num_vfs, "Number of PCI VFs to initialize");
38 static ushort rx_frag_size = 2048;
39 module_param(rx_frag_size, ushort, S_IRUGO);
40 MODULE_PARM_DESC(rx_frag_size, "Size of a fragment that holds rcvd data.");
42 static const struct pci_device_id be_dev_ids[] = {
43 { PCI_DEVICE(BE_VENDOR_ID, BE_DEVICE_ID1) },
44 { PCI_DEVICE(BE_VENDOR_ID, BE_DEVICE_ID2) },
45 { PCI_DEVICE(BE_VENDOR_ID, OC_DEVICE_ID1) },
46 { PCI_DEVICE(BE_VENDOR_ID, OC_DEVICE_ID2) },
47 { PCI_DEVICE(EMULEX_VENDOR_ID, OC_DEVICE_ID3)},
48 { PCI_DEVICE(EMULEX_VENDOR_ID, OC_DEVICE_ID4)},
49 { PCI_DEVICE(EMULEX_VENDOR_ID, OC_DEVICE_ID5)},
50 { PCI_DEVICE(EMULEX_VENDOR_ID, OC_DEVICE_ID6)},
53 MODULE_DEVICE_TABLE(pci, be_dev_ids);
54 /* UE Status Low CSR */
55 static const char * const ue_status_low_desc[] = {
90 /* UE Status High CSR */
91 static const char * const ue_status_hi_desc[] = {
126 static void be_queue_free(struct be_adapter *adapter, struct be_queue_info *q)
128 struct be_dma_mem *mem = &q->dma_mem;
131 dma_free_coherent(&adapter->pdev->dev, mem->size, mem->va,
137 static int be_queue_alloc(struct be_adapter *adapter, struct be_queue_info *q,
138 u16 len, u16 entry_size)
140 struct be_dma_mem *mem = &q->dma_mem;
142 memset(q, 0, sizeof(*q));
144 q->entry_size = entry_size;
145 mem->size = len * entry_size;
146 mem->va = dma_zalloc_coherent(&adapter->pdev->dev, mem->size, &mem->dma,
153 static void be_reg_intr_set(struct be_adapter *adapter, bool enable)
157 pci_read_config_dword(adapter->pdev, PCICFG_MEMBAR_CTRL_INT_CTRL_OFFSET,
159 enabled = reg & MEMBAR_CTRL_INT_CTRL_HOSTINTR_MASK;
161 if (!enabled && enable)
162 reg |= MEMBAR_CTRL_INT_CTRL_HOSTINTR_MASK;
163 else if (enabled && !enable)
164 reg &= ~MEMBAR_CTRL_INT_CTRL_HOSTINTR_MASK;
168 pci_write_config_dword(adapter->pdev,
169 PCICFG_MEMBAR_CTRL_INT_CTRL_OFFSET, reg);
172 static void be_intr_set(struct be_adapter *adapter, bool enable)
176 /* On lancer interrupts can't be controlled via this register */
177 if (lancer_chip(adapter))
180 if (adapter->eeh_error)
183 status = be_cmd_intr_set(adapter, enable);
185 be_reg_intr_set(adapter, enable);
188 static void be_rxq_notify(struct be_adapter *adapter, u16 qid, u16 posted)
192 val |= qid & DB_RQ_RING_ID_MASK;
193 val |= posted << DB_RQ_NUM_POSTED_SHIFT;
196 iowrite32(val, adapter->db + DB_RQ_OFFSET);
199 static void be_txq_notify(struct be_adapter *adapter, struct be_tx_obj *txo,
204 val |= txo->q.id & DB_TXULP_RING_ID_MASK;
205 val |= (posted & DB_TXULP_NUM_POSTED_MASK) << DB_TXULP_NUM_POSTED_SHIFT;
208 iowrite32(val, adapter->db + txo->db_offset);
211 static void be_eq_notify(struct be_adapter *adapter, u16 qid,
212 bool arm, bool clear_int, u16 num_popped)
216 val |= qid & DB_EQ_RING_ID_MASK;
217 val |= ((qid & DB_EQ_RING_ID_EXT_MASK) << DB_EQ_RING_ID_EXT_MASK_SHIFT);
219 if (adapter->eeh_error)
223 val |= 1 << DB_EQ_REARM_SHIFT;
225 val |= 1 << DB_EQ_CLR_SHIFT;
226 val |= 1 << DB_EQ_EVNT_SHIFT;
227 val |= num_popped << DB_EQ_NUM_POPPED_SHIFT;
228 iowrite32(val, adapter->db + DB_EQ_OFFSET);
231 void be_cq_notify(struct be_adapter *adapter, u16 qid, bool arm, u16 num_popped)
235 val |= qid & DB_CQ_RING_ID_MASK;
236 val |= ((qid & DB_CQ_RING_ID_EXT_MASK) <<
237 DB_CQ_RING_ID_EXT_MASK_SHIFT);
239 if (adapter->eeh_error)
243 val |= 1 << DB_CQ_REARM_SHIFT;
244 val |= num_popped << DB_CQ_NUM_POPPED_SHIFT;
245 iowrite32(val, adapter->db + DB_CQ_OFFSET);
248 static int be_mac_addr_set(struct net_device *netdev, void *p)
250 struct be_adapter *adapter = netdev_priv(netdev);
251 struct device *dev = &adapter->pdev->dev;
252 struct sockaddr *addr = p;
255 u32 old_pmac_id = adapter->pmac_id[0], curr_pmac_id = 0;
257 if (!is_valid_ether_addr(addr->sa_data))
258 return -EADDRNOTAVAIL;
260 /* Proceed further only if, User provided MAC is different
263 if (ether_addr_equal(addr->sa_data, netdev->dev_addr))
266 /* The PMAC_ADD cmd may fail if the VF doesn't have FILTMGMT
267 * privilege or if PF did not provision the new MAC address.
268 * On BE3, this cmd will always fail if the VF doesn't have the
269 * FILTMGMT privilege. This failure is OK, only if the PF programmed
270 * the MAC for the VF.
272 status = be_cmd_pmac_add(adapter, (u8 *)addr->sa_data,
273 adapter->if_handle, &adapter->pmac_id[0], 0);
275 curr_pmac_id = adapter->pmac_id[0];
277 /* Delete the old programmed MAC. This call may fail if the
278 * old MAC was already deleted by the PF driver.
280 if (adapter->pmac_id[0] != old_pmac_id)
281 be_cmd_pmac_del(adapter, adapter->if_handle,
285 /* Decide if the new MAC is successfully activated only after
288 status = be_cmd_get_active_mac(adapter, curr_pmac_id, mac,
289 adapter->if_handle, true, 0);
293 /* The MAC change did not happen, either due to lack of privilege
294 * or PF didn't pre-provision.
296 if (!ether_addr_equal(addr->sa_data, mac)) {
301 memcpy(netdev->dev_addr, addr->sa_data, netdev->addr_len);
302 dev_info(dev, "MAC address changed to %pM\n", mac);
305 dev_warn(dev, "MAC address change to %pM failed\n", addr->sa_data);
309 /* BE2 supports only v0 cmd */
310 static void *hw_stats_from_cmd(struct be_adapter *adapter)
312 if (BE2_chip(adapter)) {
313 struct be_cmd_resp_get_stats_v0 *cmd = adapter->stats_cmd.va;
315 return &cmd->hw_stats;
316 } else if (BE3_chip(adapter)) {
317 struct be_cmd_resp_get_stats_v1 *cmd = adapter->stats_cmd.va;
319 return &cmd->hw_stats;
321 struct be_cmd_resp_get_stats_v2 *cmd = adapter->stats_cmd.va;
323 return &cmd->hw_stats;
327 /* BE2 supports only v0 cmd */
328 static void *be_erx_stats_from_cmd(struct be_adapter *adapter)
330 if (BE2_chip(adapter)) {
331 struct be_hw_stats_v0 *hw_stats = hw_stats_from_cmd(adapter);
333 return &hw_stats->erx;
334 } else if (BE3_chip(adapter)) {
335 struct be_hw_stats_v1 *hw_stats = hw_stats_from_cmd(adapter);
337 return &hw_stats->erx;
339 struct be_hw_stats_v2 *hw_stats = hw_stats_from_cmd(adapter);
341 return &hw_stats->erx;
345 static void populate_be_v0_stats(struct be_adapter *adapter)
347 struct be_hw_stats_v0 *hw_stats = hw_stats_from_cmd(adapter);
348 struct be_pmem_stats *pmem_sts = &hw_stats->pmem;
349 struct be_rxf_stats_v0 *rxf_stats = &hw_stats->rxf;
350 struct be_port_rxf_stats_v0 *port_stats =
351 &rxf_stats->port[adapter->port_num];
352 struct be_drv_stats *drvs = &adapter->drv_stats;
354 be_dws_le_to_cpu(hw_stats, sizeof(*hw_stats));
355 drvs->rx_pause_frames = port_stats->rx_pause_frames;
356 drvs->rx_crc_errors = port_stats->rx_crc_errors;
357 drvs->rx_control_frames = port_stats->rx_control_frames;
358 drvs->rx_in_range_errors = port_stats->rx_in_range_errors;
359 drvs->rx_frame_too_long = port_stats->rx_frame_too_long;
360 drvs->rx_dropped_runt = port_stats->rx_dropped_runt;
361 drvs->rx_ip_checksum_errs = port_stats->rx_ip_checksum_errs;
362 drvs->rx_tcp_checksum_errs = port_stats->rx_tcp_checksum_errs;
363 drvs->rx_udp_checksum_errs = port_stats->rx_udp_checksum_errs;
364 drvs->rxpp_fifo_overflow_drop = port_stats->rx_fifo_overflow;
365 drvs->rx_dropped_tcp_length = port_stats->rx_dropped_tcp_length;
366 drvs->rx_dropped_too_small = port_stats->rx_dropped_too_small;
367 drvs->rx_dropped_too_short = port_stats->rx_dropped_too_short;
368 drvs->rx_out_range_errors = port_stats->rx_out_range_errors;
369 drvs->rx_input_fifo_overflow_drop = port_stats->rx_input_fifo_overflow;
370 drvs->rx_dropped_header_too_small =
371 port_stats->rx_dropped_header_too_small;
372 drvs->rx_address_filtered =
373 port_stats->rx_address_filtered +
374 port_stats->rx_vlan_filtered;
375 drvs->rx_alignment_symbol_errors =
376 port_stats->rx_alignment_symbol_errors;
378 drvs->tx_pauseframes = port_stats->tx_pauseframes;
379 drvs->tx_controlframes = port_stats->tx_controlframes;
381 if (adapter->port_num)
382 drvs->jabber_events = rxf_stats->port1_jabber_events;
384 drvs->jabber_events = rxf_stats->port0_jabber_events;
385 drvs->rx_drops_no_pbuf = rxf_stats->rx_drops_no_pbuf;
386 drvs->rx_drops_no_erx_descr = rxf_stats->rx_drops_no_erx_descr;
387 drvs->forwarded_packets = rxf_stats->forwarded_packets;
388 drvs->rx_drops_mtu = rxf_stats->rx_drops_mtu;
389 drvs->rx_drops_no_tpre_descr = rxf_stats->rx_drops_no_tpre_descr;
390 drvs->rx_drops_too_many_frags = rxf_stats->rx_drops_too_many_frags;
391 adapter->drv_stats.eth_red_drops = pmem_sts->eth_red_drops;
394 static void populate_be_v1_stats(struct be_adapter *adapter)
396 struct be_hw_stats_v1 *hw_stats = hw_stats_from_cmd(adapter);
397 struct be_pmem_stats *pmem_sts = &hw_stats->pmem;
398 struct be_rxf_stats_v1 *rxf_stats = &hw_stats->rxf;
399 struct be_port_rxf_stats_v1 *port_stats =
400 &rxf_stats->port[adapter->port_num];
401 struct be_drv_stats *drvs = &adapter->drv_stats;
403 be_dws_le_to_cpu(hw_stats, sizeof(*hw_stats));
404 drvs->pmem_fifo_overflow_drop = port_stats->pmem_fifo_overflow_drop;
405 drvs->rx_priority_pause_frames = port_stats->rx_priority_pause_frames;
406 drvs->rx_pause_frames = port_stats->rx_pause_frames;
407 drvs->rx_crc_errors = port_stats->rx_crc_errors;
408 drvs->rx_control_frames = port_stats->rx_control_frames;
409 drvs->rx_in_range_errors = port_stats->rx_in_range_errors;
410 drvs->rx_frame_too_long = port_stats->rx_frame_too_long;
411 drvs->rx_dropped_runt = port_stats->rx_dropped_runt;
412 drvs->rx_ip_checksum_errs = port_stats->rx_ip_checksum_errs;
413 drvs->rx_tcp_checksum_errs = port_stats->rx_tcp_checksum_errs;
414 drvs->rx_udp_checksum_errs = port_stats->rx_udp_checksum_errs;
415 drvs->rx_dropped_tcp_length = port_stats->rx_dropped_tcp_length;
416 drvs->rx_dropped_too_small = port_stats->rx_dropped_too_small;
417 drvs->rx_dropped_too_short = port_stats->rx_dropped_too_short;
418 drvs->rx_out_range_errors = port_stats->rx_out_range_errors;
419 drvs->rx_dropped_header_too_small =
420 port_stats->rx_dropped_header_too_small;
421 drvs->rx_input_fifo_overflow_drop =
422 port_stats->rx_input_fifo_overflow_drop;
423 drvs->rx_address_filtered = port_stats->rx_address_filtered;
424 drvs->rx_alignment_symbol_errors =
425 port_stats->rx_alignment_symbol_errors;
426 drvs->rxpp_fifo_overflow_drop = port_stats->rxpp_fifo_overflow_drop;
427 drvs->tx_pauseframes = port_stats->tx_pauseframes;
428 drvs->tx_controlframes = port_stats->tx_controlframes;
429 drvs->tx_priority_pauseframes = port_stats->tx_priority_pauseframes;
430 drvs->jabber_events = port_stats->jabber_events;
431 drvs->rx_drops_no_pbuf = rxf_stats->rx_drops_no_pbuf;
432 drvs->rx_drops_no_erx_descr = rxf_stats->rx_drops_no_erx_descr;
433 drvs->forwarded_packets = rxf_stats->forwarded_packets;
434 drvs->rx_drops_mtu = rxf_stats->rx_drops_mtu;
435 drvs->rx_drops_no_tpre_descr = rxf_stats->rx_drops_no_tpre_descr;
436 drvs->rx_drops_too_many_frags = rxf_stats->rx_drops_too_many_frags;
437 adapter->drv_stats.eth_red_drops = pmem_sts->eth_red_drops;
440 static void populate_be_v2_stats(struct be_adapter *adapter)
442 struct be_hw_stats_v2 *hw_stats = hw_stats_from_cmd(adapter);
443 struct be_pmem_stats *pmem_sts = &hw_stats->pmem;
444 struct be_rxf_stats_v2 *rxf_stats = &hw_stats->rxf;
445 struct be_port_rxf_stats_v2 *port_stats =
446 &rxf_stats->port[adapter->port_num];
447 struct be_drv_stats *drvs = &adapter->drv_stats;
449 be_dws_le_to_cpu(hw_stats, sizeof(*hw_stats));
450 drvs->pmem_fifo_overflow_drop = port_stats->pmem_fifo_overflow_drop;
451 drvs->rx_priority_pause_frames = port_stats->rx_priority_pause_frames;
452 drvs->rx_pause_frames = port_stats->rx_pause_frames;
453 drvs->rx_crc_errors = port_stats->rx_crc_errors;
454 drvs->rx_control_frames = port_stats->rx_control_frames;
455 drvs->rx_in_range_errors = port_stats->rx_in_range_errors;
456 drvs->rx_frame_too_long = port_stats->rx_frame_too_long;
457 drvs->rx_dropped_runt = port_stats->rx_dropped_runt;
458 drvs->rx_ip_checksum_errs = port_stats->rx_ip_checksum_errs;
459 drvs->rx_tcp_checksum_errs = port_stats->rx_tcp_checksum_errs;
460 drvs->rx_udp_checksum_errs = port_stats->rx_udp_checksum_errs;
461 drvs->rx_dropped_tcp_length = port_stats->rx_dropped_tcp_length;
462 drvs->rx_dropped_too_small = port_stats->rx_dropped_too_small;
463 drvs->rx_dropped_too_short = port_stats->rx_dropped_too_short;
464 drvs->rx_out_range_errors = port_stats->rx_out_range_errors;
465 drvs->rx_dropped_header_too_small =
466 port_stats->rx_dropped_header_too_small;
467 drvs->rx_input_fifo_overflow_drop =
468 port_stats->rx_input_fifo_overflow_drop;
469 drvs->rx_address_filtered = port_stats->rx_address_filtered;
470 drvs->rx_alignment_symbol_errors =
471 port_stats->rx_alignment_symbol_errors;
472 drvs->rxpp_fifo_overflow_drop = port_stats->rxpp_fifo_overflow_drop;
473 drvs->tx_pauseframes = port_stats->tx_pauseframes;
474 drvs->tx_controlframes = port_stats->tx_controlframes;
475 drvs->tx_priority_pauseframes = port_stats->tx_priority_pauseframes;
476 drvs->jabber_events = port_stats->jabber_events;
477 drvs->rx_drops_no_pbuf = rxf_stats->rx_drops_no_pbuf;
478 drvs->rx_drops_no_erx_descr = rxf_stats->rx_drops_no_erx_descr;
479 drvs->forwarded_packets = rxf_stats->forwarded_packets;
480 drvs->rx_drops_mtu = rxf_stats->rx_drops_mtu;
481 drvs->rx_drops_no_tpre_descr = rxf_stats->rx_drops_no_tpre_descr;
482 drvs->rx_drops_too_many_frags = rxf_stats->rx_drops_too_many_frags;
483 adapter->drv_stats.eth_red_drops = pmem_sts->eth_red_drops;
484 if (be_roce_supported(adapter)) {
485 drvs->rx_roce_bytes_lsd = port_stats->roce_bytes_received_lsd;
486 drvs->rx_roce_bytes_msd = port_stats->roce_bytes_received_msd;
487 drvs->rx_roce_frames = port_stats->roce_frames_received;
488 drvs->roce_drops_crc = port_stats->roce_drops_crc;
489 drvs->roce_drops_payload_len =
490 port_stats->roce_drops_payload_len;
494 static void populate_lancer_stats(struct be_adapter *adapter)
496 struct be_drv_stats *drvs = &adapter->drv_stats;
497 struct lancer_pport_stats *pport_stats = pport_stats_from_cmd(adapter);
499 be_dws_le_to_cpu(pport_stats, sizeof(*pport_stats));
500 drvs->rx_pause_frames = pport_stats->rx_pause_frames_lo;
501 drvs->rx_crc_errors = pport_stats->rx_crc_errors_lo;
502 drvs->rx_control_frames = pport_stats->rx_control_frames_lo;
503 drvs->rx_in_range_errors = pport_stats->rx_in_range_errors;
504 drvs->rx_frame_too_long = pport_stats->rx_frames_too_long_lo;
505 drvs->rx_dropped_runt = pport_stats->rx_dropped_runt;
506 drvs->rx_ip_checksum_errs = pport_stats->rx_ip_checksum_errors;
507 drvs->rx_tcp_checksum_errs = pport_stats->rx_tcp_checksum_errors;
508 drvs->rx_udp_checksum_errs = pport_stats->rx_udp_checksum_errors;
509 drvs->rx_dropped_tcp_length =
510 pport_stats->rx_dropped_invalid_tcp_length;
511 drvs->rx_dropped_too_small = pport_stats->rx_dropped_too_small;
512 drvs->rx_dropped_too_short = pport_stats->rx_dropped_too_short;
513 drvs->rx_out_range_errors = pport_stats->rx_out_of_range_errors;
514 drvs->rx_dropped_header_too_small =
515 pport_stats->rx_dropped_header_too_small;
516 drvs->rx_input_fifo_overflow_drop = pport_stats->rx_fifo_overflow;
517 drvs->rx_address_filtered =
518 pport_stats->rx_address_filtered +
519 pport_stats->rx_vlan_filtered;
520 drvs->rx_alignment_symbol_errors = pport_stats->rx_symbol_errors_lo;
521 drvs->rxpp_fifo_overflow_drop = pport_stats->rx_fifo_overflow;
522 drvs->tx_pauseframes = pport_stats->tx_pause_frames_lo;
523 drvs->tx_controlframes = pport_stats->tx_control_frames_lo;
524 drvs->jabber_events = pport_stats->rx_jabbers;
525 drvs->forwarded_packets = pport_stats->num_forwards_lo;
526 drvs->rx_drops_mtu = pport_stats->rx_drops_mtu_lo;
527 drvs->rx_drops_too_many_frags =
528 pport_stats->rx_drops_too_many_frags_lo;
531 static void accumulate_16bit_val(u32 *acc, u16 val)
533 #define lo(x) (x & 0xFFFF)
534 #define hi(x) (x & 0xFFFF0000)
535 bool wrapped = val < lo(*acc);
536 u32 newacc = hi(*acc) + val;
540 ACCESS_ONCE(*acc) = newacc;
543 static void populate_erx_stats(struct be_adapter *adapter,
544 struct be_rx_obj *rxo, u32 erx_stat)
546 if (!BEx_chip(adapter))
547 rx_stats(rxo)->rx_drops_no_frags = erx_stat;
549 /* below erx HW counter can actually wrap around after
550 * 65535. Driver accumulates a 32-bit value
552 accumulate_16bit_val(&rx_stats(rxo)->rx_drops_no_frags,
556 void be_parse_stats(struct be_adapter *adapter)
558 struct be_erx_stats_v2 *erx = be_erx_stats_from_cmd(adapter);
559 struct be_rx_obj *rxo;
563 if (lancer_chip(adapter)) {
564 populate_lancer_stats(adapter);
566 if (BE2_chip(adapter))
567 populate_be_v0_stats(adapter);
568 else if (BE3_chip(adapter))
570 populate_be_v1_stats(adapter);
572 populate_be_v2_stats(adapter);
574 /* erx_v2 is longer than v0, v1. use v2 for v0, v1 access */
575 for_all_rx_queues(adapter, rxo, i) {
576 erx_stat = erx->rx_drops_no_fragments[rxo->q.id];
577 populate_erx_stats(adapter, rxo, erx_stat);
582 static struct rtnl_link_stats64 *be_get_stats64(struct net_device *netdev,
583 struct rtnl_link_stats64 *stats)
585 struct be_adapter *adapter = netdev_priv(netdev);
586 struct be_drv_stats *drvs = &adapter->drv_stats;
587 struct be_rx_obj *rxo;
588 struct be_tx_obj *txo;
593 for_all_rx_queues(adapter, rxo, i) {
594 const struct be_rx_stats *rx_stats = rx_stats(rxo);
597 start = u64_stats_fetch_begin_irq(&rx_stats->sync);
598 pkts = rx_stats(rxo)->rx_pkts;
599 bytes = rx_stats(rxo)->rx_bytes;
600 } while (u64_stats_fetch_retry_irq(&rx_stats->sync, start));
601 stats->rx_packets += pkts;
602 stats->rx_bytes += bytes;
603 stats->multicast += rx_stats(rxo)->rx_mcast_pkts;
604 stats->rx_dropped += rx_stats(rxo)->rx_drops_no_skbs +
605 rx_stats(rxo)->rx_drops_no_frags;
608 for_all_tx_queues(adapter, txo, i) {
609 const struct be_tx_stats *tx_stats = tx_stats(txo);
612 start = u64_stats_fetch_begin_irq(&tx_stats->sync);
613 pkts = tx_stats(txo)->tx_pkts;
614 bytes = tx_stats(txo)->tx_bytes;
615 } while (u64_stats_fetch_retry_irq(&tx_stats->sync, start));
616 stats->tx_packets += pkts;
617 stats->tx_bytes += bytes;
620 /* bad pkts received */
621 stats->rx_errors = drvs->rx_crc_errors +
622 drvs->rx_alignment_symbol_errors +
623 drvs->rx_in_range_errors +
624 drvs->rx_out_range_errors +
625 drvs->rx_frame_too_long +
626 drvs->rx_dropped_too_small +
627 drvs->rx_dropped_too_short +
628 drvs->rx_dropped_header_too_small +
629 drvs->rx_dropped_tcp_length +
630 drvs->rx_dropped_runt;
632 /* detailed rx errors */
633 stats->rx_length_errors = drvs->rx_in_range_errors +
634 drvs->rx_out_range_errors +
635 drvs->rx_frame_too_long;
637 stats->rx_crc_errors = drvs->rx_crc_errors;
639 /* frame alignment errors */
640 stats->rx_frame_errors = drvs->rx_alignment_symbol_errors;
642 /* receiver fifo overrun */
643 /* drops_no_pbuf is no per i/f, it's per BE card */
644 stats->rx_fifo_errors = drvs->rxpp_fifo_overflow_drop +
645 drvs->rx_input_fifo_overflow_drop +
646 drvs->rx_drops_no_pbuf;
650 void be_link_status_update(struct be_adapter *adapter, u8 link_status)
652 struct net_device *netdev = adapter->netdev;
654 if (!(adapter->flags & BE_FLAGS_LINK_STATUS_INIT)) {
655 netif_carrier_off(netdev);
656 adapter->flags |= BE_FLAGS_LINK_STATUS_INIT;
660 netif_carrier_on(netdev);
662 netif_carrier_off(netdev);
665 static void be_tx_stats_update(struct be_tx_obj *txo, struct sk_buff *skb)
667 struct be_tx_stats *stats = tx_stats(txo);
669 u64_stats_update_begin(&stats->sync);
671 stats->tx_bytes += skb->len;
672 stats->tx_pkts += (skb_shinfo(skb)->gso_segs ? : 1);
673 u64_stats_update_end(&stats->sync);
676 /* Returns number of WRBs needed for the skb */
677 static u32 skb_wrb_cnt(struct sk_buff *skb)
679 /* +1 for the header wrb */
680 return 1 + (skb_headlen(skb) ? 1 : 0) + skb_shinfo(skb)->nr_frags;
683 static inline void wrb_fill(struct be_eth_wrb *wrb, u64 addr, int len)
685 wrb->frag_pa_hi = upper_32_bits(addr);
686 wrb->frag_pa_lo = addr & 0xFFFFFFFF;
687 wrb->frag_len = len & ETH_WRB_FRAG_LEN_MASK;
691 static inline u16 be_get_tx_vlan_tag(struct be_adapter *adapter,
697 vlan_tag = skb_vlan_tag_get(skb);
698 vlan_prio = (vlan_tag & VLAN_PRIO_MASK) >> VLAN_PRIO_SHIFT;
699 /* If vlan priority provided by OS is NOT in available bmap */
700 if (!(adapter->vlan_prio_bmap & (1 << vlan_prio)))
701 vlan_tag = (vlan_tag & ~VLAN_PRIO_MASK) |
702 adapter->recommended_prio;
707 /* Used only for IP tunnel packets */
708 static u16 skb_inner_ip_proto(struct sk_buff *skb)
710 return (inner_ip_hdr(skb)->version == 4) ?
711 inner_ip_hdr(skb)->protocol : inner_ipv6_hdr(skb)->nexthdr;
714 static u16 skb_ip_proto(struct sk_buff *skb)
716 return (ip_hdr(skb)->version == 4) ?
717 ip_hdr(skb)->protocol : ipv6_hdr(skb)->nexthdr;
720 static void wrb_fill_hdr(struct be_adapter *adapter, struct be_eth_hdr_wrb *hdr,
721 struct sk_buff *skb, u32 wrb_cnt, u32 len,
726 memset(hdr, 0, sizeof(*hdr));
728 SET_TX_WRB_HDR_BITS(crc, hdr, 1);
730 if (skb_is_gso(skb)) {
731 SET_TX_WRB_HDR_BITS(lso, hdr, 1);
732 SET_TX_WRB_HDR_BITS(lso_mss, hdr, skb_shinfo(skb)->gso_size);
733 if (skb_is_gso_v6(skb) && !lancer_chip(adapter))
734 SET_TX_WRB_HDR_BITS(lso6, hdr, 1);
735 } else if (skb->ip_summed == CHECKSUM_PARTIAL) {
736 if (skb->encapsulation) {
737 SET_TX_WRB_HDR_BITS(ipcs, hdr, 1);
738 proto = skb_inner_ip_proto(skb);
740 proto = skb_ip_proto(skb);
742 if (proto == IPPROTO_TCP)
743 SET_TX_WRB_HDR_BITS(tcpcs, hdr, 1);
744 else if (proto == IPPROTO_UDP)
745 SET_TX_WRB_HDR_BITS(udpcs, hdr, 1);
748 if (skb_vlan_tag_present(skb)) {
749 SET_TX_WRB_HDR_BITS(vlan, hdr, 1);
750 vlan_tag = be_get_tx_vlan_tag(adapter, skb);
751 SET_TX_WRB_HDR_BITS(vlan_tag, hdr, vlan_tag);
754 SET_TX_WRB_HDR_BITS(num_wrb, hdr, wrb_cnt);
755 SET_TX_WRB_HDR_BITS(len, hdr, len);
757 /* Hack to skip HW VLAN tagging needs evt = 1, compl = 0
758 * When this hack is not needed, the evt bit is set while ringing DB
761 SET_TX_WRB_HDR_BITS(event, hdr, 1);
764 static void unmap_tx_frag(struct device *dev, struct be_eth_wrb *wrb,
769 be_dws_le_to_cpu(wrb, sizeof(*wrb));
771 dma = (u64)wrb->frag_pa_hi << 32 | (u64)wrb->frag_pa_lo;
774 dma_unmap_single(dev, dma, wrb->frag_len,
777 dma_unmap_page(dev, dma, wrb->frag_len, DMA_TO_DEVICE);
781 /* Returns the number of WRBs used up by the skb */
782 static u32 be_xmit_enqueue(struct be_adapter *adapter, struct be_tx_obj *txo,
783 struct sk_buff *skb, bool skip_hw_vlan)
785 u32 i, copied = 0, wrb_cnt = skb_wrb_cnt(skb);
786 struct device *dev = &adapter->pdev->dev;
787 struct be_queue_info *txq = &txo->q;
788 struct be_eth_hdr_wrb *hdr;
789 bool map_single = false;
790 struct be_eth_wrb *wrb;
792 u16 head = txq->head;
794 hdr = queue_head_node(txq);
795 wrb_fill_hdr(adapter, hdr, skb, wrb_cnt, skb->len, skip_hw_vlan);
796 be_dws_cpu_to_le(hdr, sizeof(*hdr));
800 if (skb->len > skb->data_len) {
801 int len = skb_headlen(skb);
803 busaddr = dma_map_single(dev, skb->data, len, DMA_TO_DEVICE);
804 if (dma_mapping_error(dev, busaddr))
807 wrb = queue_head_node(txq);
808 wrb_fill(wrb, busaddr, len);
809 be_dws_cpu_to_le(wrb, sizeof(*wrb));
814 for (i = 0; i < skb_shinfo(skb)->nr_frags; i++) {
815 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i];
817 busaddr = skb_frag_dma_map(dev, frag, 0,
818 skb_frag_size(frag), DMA_TO_DEVICE);
819 if (dma_mapping_error(dev, busaddr))
821 wrb = queue_head_node(txq);
822 wrb_fill(wrb, busaddr, skb_frag_size(frag));
823 be_dws_cpu_to_le(wrb, sizeof(*wrb));
825 copied += skb_frag_size(frag);
828 BUG_ON(txo->sent_skb_list[head]);
829 txo->sent_skb_list[head] = skb;
830 txo->last_req_hdr = head;
831 atomic_add(wrb_cnt, &txq->used);
832 txo->last_req_wrb_cnt = wrb_cnt;
833 txo->pend_wrb_cnt += wrb_cnt;
835 be_tx_stats_update(txo, skb);
839 /* Bring the queue back to the state it was in before this
840 * routine was invoked.
843 /* skip the first wrb (hdr); it's not mapped */
846 wrb = queue_head_node(txq);
847 unmap_tx_frag(dev, wrb, map_single);
849 copied -= wrb->frag_len;
850 adapter->drv_stats.dma_map_errors++;
857 static struct sk_buff *be_insert_vlan_in_pkt(struct be_adapter *adapter,
863 skb = skb_share_check(skb, GFP_ATOMIC);
867 if (skb_vlan_tag_present(skb))
868 vlan_tag = be_get_tx_vlan_tag(adapter, skb);
870 if (qnq_async_evt_rcvd(adapter) && adapter->pvid) {
872 vlan_tag = adapter->pvid;
873 /* f/w workaround to set skip_hw_vlan = 1, informs the F/W to
874 * skip VLAN insertion
877 *skip_hw_vlan = true;
881 skb = vlan_insert_tag_set_proto(skb, htons(ETH_P_8021Q),
888 /* Insert the outer VLAN, if any */
889 if (adapter->qnq_vid) {
890 vlan_tag = adapter->qnq_vid;
891 skb = vlan_insert_tag_set_proto(skb, htons(ETH_P_8021Q),
896 *skip_hw_vlan = true;
902 static bool be_ipv6_exthdr_check(struct sk_buff *skb)
904 struct ethhdr *eh = (struct ethhdr *)skb->data;
905 u16 offset = ETH_HLEN;
907 if (eh->h_proto == htons(ETH_P_IPV6)) {
908 struct ipv6hdr *ip6h = (struct ipv6hdr *)(skb->data + offset);
910 offset += sizeof(struct ipv6hdr);
911 if (ip6h->nexthdr != NEXTHDR_TCP &&
912 ip6h->nexthdr != NEXTHDR_UDP) {
913 struct ipv6_opt_hdr *ehdr =
914 (struct ipv6_opt_hdr *)(skb->data + offset);
916 /* offending pkt: 2nd byte following IPv6 hdr is 0xff */
917 if (ehdr->hdrlen == 0xff)
924 static int be_vlan_tag_tx_chk(struct be_adapter *adapter, struct sk_buff *skb)
926 return skb_vlan_tag_present(skb) || adapter->pvid || adapter->qnq_vid;
929 static int be_ipv6_tx_stall_chk(struct be_adapter *adapter, struct sk_buff *skb)
931 return BE3_chip(adapter) && be_ipv6_exthdr_check(skb);
934 static struct sk_buff *be_lancer_xmit_workarounds(struct be_adapter *adapter,
938 struct vlan_ethhdr *veh = (struct vlan_ethhdr *)skb->data;
939 unsigned int eth_hdr_len;
942 /* For padded packets, BE HW modifies tot_len field in IP header
943 * incorrecly when VLAN tag is inserted by HW.
944 * For padded packets, Lancer computes incorrect checksum.
946 eth_hdr_len = ntohs(skb->protocol) == ETH_P_8021Q ?
947 VLAN_ETH_HLEN : ETH_HLEN;
948 if (skb->len <= 60 &&
949 (lancer_chip(adapter) || skb_vlan_tag_present(skb)) &&
951 ip = (struct iphdr *)ip_hdr(skb);
952 pskb_trim(skb, eth_hdr_len + ntohs(ip->tot_len));
955 /* If vlan tag is already inlined in the packet, skip HW VLAN
956 * tagging in pvid-tagging mode
958 if (be_pvid_tagging_enabled(adapter) &&
959 veh->h_vlan_proto == htons(ETH_P_8021Q))
960 *skip_hw_vlan = true;
962 /* HW has a bug wherein it will calculate CSUM for VLAN
963 * pkts even though it is disabled.
964 * Manually insert VLAN in pkt.
966 if (skb->ip_summed != CHECKSUM_PARTIAL &&
967 skb_vlan_tag_present(skb)) {
968 skb = be_insert_vlan_in_pkt(adapter, skb, skip_hw_vlan);
973 /* HW may lockup when VLAN HW tagging is requested on
974 * certain ipv6 packets. Drop such pkts if the HW workaround to
975 * skip HW tagging is not enabled by FW.
977 if (unlikely(be_ipv6_tx_stall_chk(adapter, skb) &&
978 (adapter->pvid || adapter->qnq_vid) &&
979 !qnq_async_evt_rcvd(adapter)))
982 /* Manual VLAN tag insertion to prevent:
983 * ASIC lockup when the ASIC inserts VLAN tag into
984 * certain ipv6 packets. Insert VLAN tags in driver,
985 * and set event, completion, vlan bits accordingly
988 if (be_ipv6_tx_stall_chk(adapter, skb) &&
989 be_vlan_tag_tx_chk(adapter, skb)) {
990 skb = be_insert_vlan_in_pkt(adapter, skb, skip_hw_vlan);
997 dev_kfree_skb_any(skb);
1002 static struct sk_buff *be_xmit_workarounds(struct be_adapter *adapter,
1003 struct sk_buff *skb,
1006 /* Lancer, SH-R ASICs have a bug wherein Packets that are 32 bytes or
1007 * less may cause a transmit stall on that port. So the work-around is
1008 * to pad short packets (<= 32 bytes) to a 36-byte length.
1010 if (unlikely(!BEx_chip(adapter) && skb->len <= 32)) {
1011 if (skb_put_padto(skb, 36))
1015 if (BEx_chip(adapter) || lancer_chip(adapter)) {
1016 skb = be_lancer_xmit_workarounds(adapter, skb, skip_hw_vlan);
1024 static void be_xmit_flush(struct be_adapter *adapter, struct be_tx_obj *txo)
1026 struct be_queue_info *txq = &txo->q;
1027 struct be_eth_hdr_wrb *hdr = queue_index_node(txq, txo->last_req_hdr);
1029 /* Mark the last request eventable if it hasn't been marked already */
1030 if (!(hdr->dw[2] & cpu_to_le32(TX_HDR_WRB_EVT)))
1031 hdr->dw[2] |= cpu_to_le32(TX_HDR_WRB_EVT | TX_HDR_WRB_COMPL);
1033 /* compose a dummy wrb if there are odd set of wrbs to notify */
1034 if (!lancer_chip(adapter) && (txo->pend_wrb_cnt & 1)) {
1035 wrb_fill(queue_head_node(txq), 0, 0);
1036 queue_head_inc(txq);
1037 atomic_inc(&txq->used);
1038 txo->pend_wrb_cnt++;
1039 hdr->dw[2] &= ~cpu_to_le32(TX_HDR_WRB_NUM_MASK <<
1040 TX_HDR_WRB_NUM_SHIFT);
1041 hdr->dw[2] |= cpu_to_le32((txo->last_req_wrb_cnt + 1) <<
1042 TX_HDR_WRB_NUM_SHIFT);
1044 be_txq_notify(adapter, txo, txo->pend_wrb_cnt);
1045 txo->pend_wrb_cnt = 0;
1048 static netdev_tx_t be_xmit(struct sk_buff *skb, struct net_device *netdev)
1050 bool skip_hw_vlan = false, flush = !skb->xmit_more;
1051 struct be_adapter *adapter = netdev_priv(netdev);
1052 u16 q_idx = skb_get_queue_mapping(skb);
1053 struct be_tx_obj *txo = &adapter->tx_obj[q_idx];
1054 struct be_queue_info *txq = &txo->q;
1057 skb = be_xmit_workarounds(adapter, skb, &skip_hw_vlan);
1061 wrb_cnt = be_xmit_enqueue(adapter, txo, skb, skip_hw_vlan);
1062 if (unlikely(!wrb_cnt)) {
1063 dev_kfree_skb_any(skb);
1067 if ((atomic_read(&txq->used) + BE_MAX_TX_FRAG_COUNT) >= txq->len) {
1068 netif_stop_subqueue(netdev, q_idx);
1069 tx_stats(txo)->tx_stops++;
1072 if (flush || __netif_subqueue_stopped(netdev, q_idx))
1073 be_xmit_flush(adapter, txo);
1075 return NETDEV_TX_OK;
1077 tx_stats(txo)->tx_drv_drops++;
1078 /* Flush the already enqueued tx requests */
1079 if (flush && txo->pend_wrb_cnt)
1080 be_xmit_flush(adapter, txo);
1082 return NETDEV_TX_OK;
1085 static int be_change_mtu(struct net_device *netdev, int new_mtu)
1087 struct be_adapter *adapter = netdev_priv(netdev);
1088 struct device *dev = &adapter->pdev->dev;
1090 if (new_mtu < BE_MIN_MTU || new_mtu > BE_MAX_MTU) {
1091 dev_info(dev, "MTU must be between %d and %d bytes\n",
1092 BE_MIN_MTU, BE_MAX_MTU);
1096 dev_info(dev, "MTU changed from %d to %d bytes\n",
1097 netdev->mtu, new_mtu);
1098 netdev->mtu = new_mtu;
1103 * A max of 64 (BE_NUM_VLANS_SUPPORTED) vlans can be configured in BE.
1104 * If the user configures more, place BE in vlan promiscuous mode.
1106 static int be_vid_config(struct be_adapter *adapter)
1108 struct device *dev = &adapter->pdev->dev;
1109 u16 vids[BE_NUM_VLANS_SUPPORTED];
1113 /* No need to further configure vids if in promiscuous mode */
1114 if (adapter->promiscuous)
1117 if (adapter->vlans_added > be_max_vlans(adapter))
1118 goto set_vlan_promisc;
1120 /* Construct VLAN Table to give to HW */
1121 for_each_set_bit(i, adapter->vids, VLAN_N_VID)
1122 vids[num++] = cpu_to_le16(i);
1124 status = be_cmd_vlan_config(adapter, adapter->if_handle, vids, num);
1126 /* Set to VLAN promisc mode as setting VLAN filter failed */
1127 if (addl_status(status) ==
1128 MCC_ADDL_STATUS_INSUFFICIENT_RESOURCES)
1129 goto set_vlan_promisc;
1130 dev_err(dev, "Setting HW VLAN filtering failed\n");
1132 if (adapter->flags & BE_FLAGS_VLAN_PROMISC) {
1133 /* hw VLAN filtering re-enabled. */
1134 status = be_cmd_rx_filter(adapter,
1135 BE_FLAGS_VLAN_PROMISC, OFF);
1138 "Disabling VLAN Promiscuous mode\n");
1139 adapter->flags &= ~BE_FLAGS_VLAN_PROMISC;
1147 if (adapter->flags & BE_FLAGS_VLAN_PROMISC)
1150 status = be_cmd_rx_filter(adapter, BE_FLAGS_VLAN_PROMISC, ON);
1152 dev_info(dev, "Enable VLAN Promiscuous mode\n");
1153 adapter->flags |= BE_FLAGS_VLAN_PROMISC;
1155 dev_err(dev, "Failed to enable VLAN Promiscuous mode\n");
1159 static int be_vlan_add_vid(struct net_device *netdev, __be16 proto, u16 vid)
1161 struct be_adapter *adapter = netdev_priv(netdev);
1164 /* Packets with VID 0 are always received by Lancer by default */
1165 if (lancer_chip(adapter) && vid == 0)
1168 if (test_bit(vid, adapter->vids))
1171 set_bit(vid, adapter->vids);
1172 adapter->vlans_added++;
1174 status = be_vid_config(adapter);
1176 adapter->vlans_added--;
1177 clear_bit(vid, adapter->vids);
1183 static int be_vlan_rem_vid(struct net_device *netdev, __be16 proto, u16 vid)
1185 struct be_adapter *adapter = netdev_priv(netdev);
1187 /* Packets with VID 0 are always received by Lancer by default */
1188 if (lancer_chip(adapter) && vid == 0)
1191 clear_bit(vid, adapter->vids);
1192 adapter->vlans_added--;
1194 return be_vid_config(adapter);
1197 static void be_clear_promisc(struct be_adapter *adapter)
1199 adapter->promiscuous = false;
1200 adapter->flags &= ~(BE_FLAGS_VLAN_PROMISC | BE_FLAGS_MCAST_PROMISC);
1202 be_cmd_rx_filter(adapter, IFF_PROMISC, OFF);
1205 static void be_set_rx_mode(struct net_device *netdev)
1207 struct be_adapter *adapter = netdev_priv(netdev);
1210 if (netdev->flags & IFF_PROMISC) {
1211 be_cmd_rx_filter(adapter, IFF_PROMISC, ON);
1212 adapter->promiscuous = true;
1216 /* BE was previously in promiscuous mode; disable it */
1217 if (adapter->promiscuous) {
1218 be_clear_promisc(adapter);
1219 if (adapter->vlans_added)
1220 be_vid_config(adapter);
1223 /* Enable multicast promisc if num configured exceeds what we support */
1224 if (netdev->flags & IFF_ALLMULTI ||
1225 netdev_mc_count(netdev) > be_max_mc(adapter))
1226 goto set_mcast_promisc;
1228 if (netdev_uc_count(netdev) != adapter->uc_macs) {
1229 struct netdev_hw_addr *ha;
1230 int i = 1; /* First slot is claimed by the Primary MAC */
1232 for (; adapter->uc_macs > 0; adapter->uc_macs--, i++) {
1233 be_cmd_pmac_del(adapter, adapter->if_handle,
1234 adapter->pmac_id[i], 0);
1237 if (netdev_uc_count(netdev) > be_max_uc(adapter)) {
1238 be_cmd_rx_filter(adapter, IFF_PROMISC, ON);
1239 adapter->promiscuous = true;
1243 netdev_for_each_uc_addr(ha, adapter->netdev) {
1244 adapter->uc_macs++; /* First slot is for Primary MAC */
1245 be_cmd_pmac_add(adapter, (u8 *)ha->addr,
1247 &adapter->pmac_id[adapter->uc_macs], 0);
1251 status = be_cmd_rx_filter(adapter, IFF_MULTICAST, ON);
1253 if (adapter->flags & BE_FLAGS_MCAST_PROMISC)
1254 adapter->flags &= ~BE_FLAGS_MCAST_PROMISC;
1259 if (adapter->flags & BE_FLAGS_MCAST_PROMISC)
1262 /* Set to MCAST promisc mode if setting MULTICAST address fails
1263 * or if num configured exceeds what we support
1265 status = be_cmd_rx_filter(adapter, IFF_ALLMULTI, ON);
1267 adapter->flags |= BE_FLAGS_MCAST_PROMISC;
1272 static int be_set_vf_mac(struct net_device *netdev, int vf, u8 *mac)
1274 struct be_adapter *adapter = netdev_priv(netdev);
1275 struct be_vf_cfg *vf_cfg = &adapter->vf_cfg[vf];
1278 if (!sriov_enabled(adapter))
1281 if (!is_valid_ether_addr(mac) || vf >= adapter->num_vfs)
1284 /* Proceed further only if user provided MAC is different
1287 if (ether_addr_equal(mac, vf_cfg->mac_addr))
1290 if (BEx_chip(adapter)) {
1291 be_cmd_pmac_del(adapter, vf_cfg->if_handle, vf_cfg->pmac_id,
1294 status = be_cmd_pmac_add(adapter, mac, vf_cfg->if_handle,
1295 &vf_cfg->pmac_id, vf + 1);
1297 status = be_cmd_set_mac(adapter, mac, vf_cfg->if_handle,
1302 dev_err(&adapter->pdev->dev, "MAC %pM set on VF %d Failed: %#x",
1304 return be_cmd_status(status);
1307 ether_addr_copy(vf_cfg->mac_addr, mac);
1312 static int be_get_vf_config(struct net_device *netdev, int vf,
1313 struct ifla_vf_info *vi)
1315 struct be_adapter *adapter = netdev_priv(netdev);
1316 struct be_vf_cfg *vf_cfg = &adapter->vf_cfg[vf];
1318 if (!sriov_enabled(adapter))
1321 if (vf >= adapter->num_vfs)
1325 vi->max_tx_rate = vf_cfg->tx_rate;
1326 vi->min_tx_rate = 0;
1327 vi->vlan = vf_cfg->vlan_tag & VLAN_VID_MASK;
1328 vi->qos = vf_cfg->vlan_tag >> VLAN_PRIO_SHIFT;
1329 memcpy(&vi->mac, vf_cfg->mac_addr, ETH_ALEN);
1330 vi->linkstate = adapter->vf_cfg[vf].plink_tracking;
1335 static int be_set_vf_vlan(struct net_device *netdev, int vf, u16 vlan, u8 qos)
1337 struct be_adapter *adapter = netdev_priv(netdev);
1338 struct be_vf_cfg *vf_cfg = &adapter->vf_cfg[vf];
1341 if (!sriov_enabled(adapter))
1344 if (vf >= adapter->num_vfs || vlan > 4095 || qos > 7)
1348 vlan |= qos << VLAN_PRIO_SHIFT;
1349 if (vf_cfg->vlan_tag != vlan)
1350 status = be_cmd_set_hsw_config(adapter, vlan, vf + 1,
1351 vf_cfg->if_handle, 0);
1353 /* Reset Transparent Vlan Tagging. */
1354 status = be_cmd_set_hsw_config(adapter, BE_RESET_VLAN_TAG_ID,
1355 vf + 1, vf_cfg->if_handle, 0);
1359 dev_err(&adapter->pdev->dev,
1360 "VLAN %d config on VF %d failed : %#x\n", vlan,
1362 return be_cmd_status(status);
1365 vf_cfg->vlan_tag = vlan;
1370 static int be_set_vf_tx_rate(struct net_device *netdev, int vf,
1371 int min_tx_rate, int max_tx_rate)
1373 struct be_adapter *adapter = netdev_priv(netdev);
1374 struct device *dev = &adapter->pdev->dev;
1375 int percent_rate, status = 0;
1379 if (!sriov_enabled(adapter))
1382 if (vf >= adapter->num_vfs)
1391 status = be_cmd_link_status_query(adapter, &link_speed,
1397 dev_err(dev, "TX-rate setting not allowed when link is down\n");
1402 if (max_tx_rate < 100 || max_tx_rate > link_speed) {
1403 dev_err(dev, "TX-rate must be between 100 and %d Mbps\n",
1409 /* On Skyhawk the QOS setting must be done only as a % value */
1410 percent_rate = link_speed / 100;
1411 if (skyhawk_chip(adapter) && (max_tx_rate % percent_rate)) {
1412 dev_err(dev, "TX-rate must be a multiple of %d Mbps\n",
1419 status = be_cmd_config_qos(adapter, max_tx_rate, link_speed, vf + 1);
1423 adapter->vf_cfg[vf].tx_rate = max_tx_rate;
1427 dev_err(dev, "TX-rate setting of %dMbps on VF%d failed\n",
1429 return be_cmd_status(status);
1432 static int be_set_vf_link_state(struct net_device *netdev, int vf,
1435 struct be_adapter *adapter = netdev_priv(netdev);
1438 if (!sriov_enabled(adapter))
1441 if (vf >= adapter->num_vfs)
1444 status = be_cmd_set_logical_link_config(adapter, link_state, vf+1);
1446 dev_err(&adapter->pdev->dev,
1447 "Link state change on VF %d failed: %#x\n", vf, status);
1448 return be_cmd_status(status);
1451 adapter->vf_cfg[vf].plink_tracking = link_state;
1456 static void be_aic_update(struct be_aic_obj *aic, u64 rx_pkts, u64 tx_pkts,
1459 aic->rx_pkts_prev = rx_pkts;
1460 aic->tx_reqs_prev = tx_pkts;
1464 static void be_eqd_update(struct be_adapter *adapter)
1466 struct be_set_eqd set_eqd[MAX_EVT_QS];
1467 int eqd, i, num = 0, start;
1468 struct be_aic_obj *aic;
1469 struct be_eq_obj *eqo;
1470 struct be_rx_obj *rxo;
1471 struct be_tx_obj *txo;
1472 u64 rx_pkts, tx_pkts;
1476 for_all_evt_queues(adapter, eqo, i) {
1477 aic = &adapter->aic_obj[eqo->idx];
1485 rxo = &adapter->rx_obj[eqo->idx];
1487 start = u64_stats_fetch_begin_irq(&rxo->stats.sync);
1488 rx_pkts = rxo->stats.rx_pkts;
1489 } while (u64_stats_fetch_retry_irq(&rxo->stats.sync, start));
1491 txo = &adapter->tx_obj[eqo->idx];
1493 start = u64_stats_fetch_begin_irq(&txo->stats.sync);
1494 tx_pkts = txo->stats.tx_reqs;
1495 } while (u64_stats_fetch_retry_irq(&txo->stats.sync, start));
1497 /* Skip, if wrapped around or first calculation */
1499 if (!aic->jiffies || time_before(now, aic->jiffies) ||
1500 rx_pkts < aic->rx_pkts_prev ||
1501 tx_pkts < aic->tx_reqs_prev) {
1502 be_aic_update(aic, rx_pkts, tx_pkts, now);
1506 delta = jiffies_to_msecs(now - aic->jiffies);
1507 pps = (((u32)(rx_pkts - aic->rx_pkts_prev) * 1000) / delta) +
1508 (((u32)(tx_pkts - aic->tx_reqs_prev) * 1000) / delta);
1509 eqd = (pps / 15000) << 2;
1513 eqd = min_t(u32, eqd, aic->max_eqd);
1514 eqd = max_t(u32, eqd, aic->min_eqd);
1516 be_aic_update(aic, rx_pkts, tx_pkts, now);
1518 if (eqd != aic->prev_eqd) {
1519 set_eqd[num].delay_multiplier = (eqd * 65)/100;
1520 set_eqd[num].eq_id = eqo->q.id;
1521 aic->prev_eqd = eqd;
1527 be_cmd_modify_eqd(adapter, set_eqd, num);
1530 static void be_rx_stats_update(struct be_rx_obj *rxo,
1531 struct be_rx_compl_info *rxcp)
1533 struct be_rx_stats *stats = rx_stats(rxo);
1535 u64_stats_update_begin(&stats->sync);
1537 stats->rx_bytes += rxcp->pkt_size;
1539 if (rxcp->pkt_type == BE_MULTICAST_PACKET)
1540 stats->rx_mcast_pkts++;
1542 stats->rx_compl_err++;
1543 u64_stats_update_end(&stats->sync);
1546 static inline bool csum_passed(struct be_rx_compl_info *rxcp)
1548 /* L4 checksum is not reliable for non TCP/UDP packets.
1549 * Also ignore ipcksm for ipv6 pkts
1551 return (rxcp->tcpf || rxcp->udpf) && rxcp->l4_csum &&
1552 (rxcp->ip_csum || rxcp->ipv6) && !rxcp->err;
1555 static struct be_rx_page_info *get_rx_page_info(struct be_rx_obj *rxo)
1557 struct be_adapter *adapter = rxo->adapter;
1558 struct be_rx_page_info *rx_page_info;
1559 struct be_queue_info *rxq = &rxo->q;
1560 u16 frag_idx = rxq->tail;
1562 rx_page_info = &rxo->page_info_tbl[frag_idx];
1563 BUG_ON(!rx_page_info->page);
1565 if (rx_page_info->last_frag) {
1566 dma_unmap_page(&adapter->pdev->dev,
1567 dma_unmap_addr(rx_page_info, bus),
1568 adapter->big_page_size, DMA_FROM_DEVICE);
1569 rx_page_info->last_frag = false;
1571 dma_sync_single_for_cpu(&adapter->pdev->dev,
1572 dma_unmap_addr(rx_page_info, bus),
1573 rx_frag_size, DMA_FROM_DEVICE);
1576 queue_tail_inc(rxq);
1577 atomic_dec(&rxq->used);
1578 return rx_page_info;
1581 /* Throwaway the data in the Rx completion */
1582 static void be_rx_compl_discard(struct be_rx_obj *rxo,
1583 struct be_rx_compl_info *rxcp)
1585 struct be_rx_page_info *page_info;
1586 u16 i, num_rcvd = rxcp->num_rcvd;
1588 for (i = 0; i < num_rcvd; i++) {
1589 page_info = get_rx_page_info(rxo);
1590 put_page(page_info->page);
1591 memset(page_info, 0, sizeof(*page_info));
1596 * skb_fill_rx_data forms a complete skb for an ether frame
1597 * indicated by rxcp.
1599 static void skb_fill_rx_data(struct be_rx_obj *rxo, struct sk_buff *skb,
1600 struct be_rx_compl_info *rxcp)
1602 struct be_rx_page_info *page_info;
1604 u16 hdr_len, curr_frag_len, remaining;
1607 page_info = get_rx_page_info(rxo);
1608 start = page_address(page_info->page) + page_info->page_offset;
1611 /* Copy data in the first descriptor of this completion */
1612 curr_frag_len = min(rxcp->pkt_size, rx_frag_size);
1614 skb->len = curr_frag_len;
1615 if (curr_frag_len <= BE_HDR_LEN) { /* tiny packet */
1616 memcpy(skb->data, start, curr_frag_len);
1617 /* Complete packet has now been moved to data */
1618 put_page(page_info->page);
1620 skb->tail += curr_frag_len;
1623 memcpy(skb->data, start, hdr_len);
1624 skb_shinfo(skb)->nr_frags = 1;
1625 skb_frag_set_page(skb, 0, page_info->page);
1626 skb_shinfo(skb)->frags[0].page_offset =
1627 page_info->page_offset + hdr_len;
1628 skb_frag_size_set(&skb_shinfo(skb)->frags[0],
1629 curr_frag_len - hdr_len);
1630 skb->data_len = curr_frag_len - hdr_len;
1631 skb->truesize += rx_frag_size;
1632 skb->tail += hdr_len;
1634 page_info->page = NULL;
1636 if (rxcp->pkt_size <= rx_frag_size) {
1637 BUG_ON(rxcp->num_rcvd != 1);
1641 /* More frags present for this completion */
1642 remaining = rxcp->pkt_size - curr_frag_len;
1643 for (i = 1, j = 0; i < rxcp->num_rcvd; i++) {
1644 page_info = get_rx_page_info(rxo);
1645 curr_frag_len = min(remaining, rx_frag_size);
1647 /* Coalesce all frags from the same physical page in one slot */
1648 if (page_info->page_offset == 0) {
1651 skb_frag_set_page(skb, j, page_info->page);
1652 skb_shinfo(skb)->frags[j].page_offset =
1653 page_info->page_offset;
1654 skb_frag_size_set(&skb_shinfo(skb)->frags[j], 0);
1655 skb_shinfo(skb)->nr_frags++;
1657 put_page(page_info->page);
1660 skb_frag_size_add(&skb_shinfo(skb)->frags[j], curr_frag_len);
1661 skb->len += curr_frag_len;
1662 skb->data_len += curr_frag_len;
1663 skb->truesize += rx_frag_size;
1664 remaining -= curr_frag_len;
1665 page_info->page = NULL;
1667 BUG_ON(j > MAX_SKB_FRAGS);
1670 /* Process the RX completion indicated by rxcp when GRO is disabled */
1671 static void be_rx_compl_process(struct be_rx_obj *rxo, struct napi_struct *napi,
1672 struct be_rx_compl_info *rxcp)
1674 struct be_adapter *adapter = rxo->adapter;
1675 struct net_device *netdev = adapter->netdev;
1676 struct sk_buff *skb;
1678 skb = netdev_alloc_skb_ip_align(netdev, BE_RX_SKB_ALLOC_SIZE);
1679 if (unlikely(!skb)) {
1680 rx_stats(rxo)->rx_drops_no_skbs++;
1681 be_rx_compl_discard(rxo, rxcp);
1685 skb_fill_rx_data(rxo, skb, rxcp);
1687 if (likely((netdev->features & NETIF_F_RXCSUM) && csum_passed(rxcp)))
1688 skb->ip_summed = CHECKSUM_UNNECESSARY;
1690 skb_checksum_none_assert(skb);
1692 skb->protocol = eth_type_trans(skb, netdev);
1693 skb_record_rx_queue(skb, rxo - &adapter->rx_obj[0]);
1694 if (netdev->features & NETIF_F_RXHASH)
1695 skb_set_hash(skb, rxcp->rss_hash, PKT_HASH_TYPE_L3);
1697 skb->csum_level = rxcp->tunneled;
1698 skb_mark_napi_id(skb, napi);
1701 __vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q), rxcp->vlan_tag);
1703 netif_receive_skb(skb);
1706 /* Process the RX completion indicated by rxcp when GRO is enabled */
1707 static void be_rx_compl_process_gro(struct be_rx_obj *rxo,
1708 struct napi_struct *napi,
1709 struct be_rx_compl_info *rxcp)
1711 struct be_adapter *adapter = rxo->adapter;
1712 struct be_rx_page_info *page_info;
1713 struct sk_buff *skb = NULL;
1714 u16 remaining, curr_frag_len;
1717 skb = napi_get_frags(napi);
1719 be_rx_compl_discard(rxo, rxcp);
1723 remaining = rxcp->pkt_size;
1724 for (i = 0, j = -1; i < rxcp->num_rcvd; i++) {
1725 page_info = get_rx_page_info(rxo);
1727 curr_frag_len = min(remaining, rx_frag_size);
1729 /* Coalesce all frags from the same physical page in one slot */
1730 if (i == 0 || page_info->page_offset == 0) {
1731 /* First frag or Fresh page */
1733 skb_frag_set_page(skb, j, page_info->page);
1734 skb_shinfo(skb)->frags[j].page_offset =
1735 page_info->page_offset;
1736 skb_frag_size_set(&skb_shinfo(skb)->frags[j], 0);
1738 put_page(page_info->page);
1740 skb_frag_size_add(&skb_shinfo(skb)->frags[j], curr_frag_len);
1741 skb->truesize += rx_frag_size;
1742 remaining -= curr_frag_len;
1743 memset(page_info, 0, sizeof(*page_info));
1745 BUG_ON(j > MAX_SKB_FRAGS);
1747 skb_shinfo(skb)->nr_frags = j + 1;
1748 skb->len = rxcp->pkt_size;
1749 skb->data_len = rxcp->pkt_size;
1750 skb->ip_summed = CHECKSUM_UNNECESSARY;
1751 skb_record_rx_queue(skb, rxo - &adapter->rx_obj[0]);
1752 if (adapter->netdev->features & NETIF_F_RXHASH)
1753 skb_set_hash(skb, rxcp->rss_hash, PKT_HASH_TYPE_L3);
1755 skb->csum_level = rxcp->tunneled;
1756 skb_mark_napi_id(skb, napi);
1759 __vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q), rxcp->vlan_tag);
1761 napi_gro_frags(napi);
1764 static void be_parse_rx_compl_v1(struct be_eth_rx_compl *compl,
1765 struct be_rx_compl_info *rxcp)
1767 rxcp->pkt_size = GET_RX_COMPL_V1_BITS(pktsize, compl);
1768 rxcp->vlanf = GET_RX_COMPL_V1_BITS(vtp, compl);
1769 rxcp->err = GET_RX_COMPL_V1_BITS(err, compl);
1770 rxcp->tcpf = GET_RX_COMPL_V1_BITS(tcpf, compl);
1771 rxcp->udpf = GET_RX_COMPL_V1_BITS(udpf, compl);
1772 rxcp->ip_csum = GET_RX_COMPL_V1_BITS(ipcksm, compl);
1773 rxcp->l4_csum = GET_RX_COMPL_V1_BITS(l4_cksm, compl);
1774 rxcp->ipv6 = GET_RX_COMPL_V1_BITS(ip_version, compl);
1775 rxcp->num_rcvd = GET_RX_COMPL_V1_BITS(numfrags, compl);
1776 rxcp->pkt_type = GET_RX_COMPL_V1_BITS(cast_enc, compl);
1777 rxcp->rss_hash = GET_RX_COMPL_V1_BITS(rsshash, compl);
1779 rxcp->qnq = GET_RX_COMPL_V1_BITS(qnq, compl);
1780 rxcp->vlan_tag = GET_RX_COMPL_V1_BITS(vlan_tag, compl);
1782 rxcp->port = GET_RX_COMPL_V1_BITS(port, compl);
1784 GET_RX_COMPL_V1_BITS(tunneled, compl);
1787 static void be_parse_rx_compl_v0(struct be_eth_rx_compl *compl,
1788 struct be_rx_compl_info *rxcp)
1790 rxcp->pkt_size = GET_RX_COMPL_V0_BITS(pktsize, compl);
1791 rxcp->vlanf = GET_RX_COMPL_V0_BITS(vtp, compl);
1792 rxcp->err = GET_RX_COMPL_V0_BITS(err, compl);
1793 rxcp->tcpf = GET_RX_COMPL_V0_BITS(tcpf, compl);
1794 rxcp->udpf = GET_RX_COMPL_V0_BITS(udpf, compl);
1795 rxcp->ip_csum = GET_RX_COMPL_V0_BITS(ipcksm, compl);
1796 rxcp->l4_csum = GET_RX_COMPL_V0_BITS(l4_cksm, compl);
1797 rxcp->ipv6 = GET_RX_COMPL_V0_BITS(ip_version, compl);
1798 rxcp->num_rcvd = GET_RX_COMPL_V0_BITS(numfrags, compl);
1799 rxcp->pkt_type = GET_RX_COMPL_V0_BITS(cast_enc, compl);
1800 rxcp->rss_hash = GET_RX_COMPL_V0_BITS(rsshash, compl);
1802 rxcp->qnq = GET_RX_COMPL_V0_BITS(qnq, compl);
1803 rxcp->vlan_tag = GET_RX_COMPL_V0_BITS(vlan_tag, compl);
1805 rxcp->port = GET_RX_COMPL_V0_BITS(port, compl);
1806 rxcp->ip_frag = GET_RX_COMPL_V0_BITS(ip_frag, compl);
1809 static struct be_rx_compl_info *be_rx_compl_get(struct be_rx_obj *rxo)
1811 struct be_eth_rx_compl *compl = queue_tail_node(&rxo->cq);
1812 struct be_rx_compl_info *rxcp = &rxo->rxcp;
1813 struct be_adapter *adapter = rxo->adapter;
1815 /* For checking the valid bit it is Ok to use either definition as the
1816 * valid bit is at the same position in both v0 and v1 Rx compl */
1817 if (compl->dw[offsetof(struct amap_eth_rx_compl_v1, valid) / 32] == 0)
1821 be_dws_le_to_cpu(compl, sizeof(*compl));
1823 if (adapter->be3_native)
1824 be_parse_rx_compl_v1(compl, rxcp);
1826 be_parse_rx_compl_v0(compl, rxcp);
1832 /* In QNQ modes, if qnq bit is not set, then the packet was
1833 * tagged only with the transparent outer vlan-tag and must
1834 * not be treated as a vlan packet by host
1836 if (be_is_qnq_mode(adapter) && !rxcp->qnq)
1839 if (!lancer_chip(adapter))
1840 rxcp->vlan_tag = swab16(rxcp->vlan_tag);
1842 if (adapter->pvid == (rxcp->vlan_tag & VLAN_VID_MASK) &&
1843 !test_bit(rxcp->vlan_tag, adapter->vids))
1847 /* As the compl has been parsed, reset it; we wont touch it again */
1848 compl->dw[offsetof(struct amap_eth_rx_compl_v1, valid) / 32] = 0;
1850 queue_tail_inc(&rxo->cq);
1854 static inline struct page *be_alloc_pages(u32 size, gfp_t gfp)
1856 u32 order = get_order(size);
1860 return alloc_pages(gfp, order);
1864 * Allocate a page, split it to fragments of size rx_frag_size and post as
1865 * receive buffers to BE
1867 static void be_post_rx_frags(struct be_rx_obj *rxo, gfp_t gfp, u32 frags_needed)
1869 struct be_adapter *adapter = rxo->adapter;
1870 struct be_rx_page_info *page_info = NULL, *prev_page_info = NULL;
1871 struct be_queue_info *rxq = &rxo->q;
1872 struct page *pagep = NULL;
1873 struct device *dev = &adapter->pdev->dev;
1874 struct be_eth_rx_d *rxd;
1875 u64 page_dmaaddr = 0, frag_dmaaddr;
1876 u32 posted, page_offset = 0, notify = 0;
1878 page_info = &rxo->page_info_tbl[rxq->head];
1879 for (posted = 0; posted < frags_needed && !page_info->page; posted++) {
1881 pagep = be_alloc_pages(adapter->big_page_size, gfp);
1882 if (unlikely(!pagep)) {
1883 rx_stats(rxo)->rx_post_fail++;
1886 page_dmaaddr = dma_map_page(dev, pagep, 0,
1887 adapter->big_page_size,
1889 if (dma_mapping_error(dev, page_dmaaddr)) {
1892 adapter->drv_stats.dma_map_errors++;
1898 page_offset += rx_frag_size;
1900 page_info->page_offset = page_offset;
1901 page_info->page = pagep;
1903 rxd = queue_head_node(rxq);
1904 frag_dmaaddr = page_dmaaddr + page_info->page_offset;
1905 rxd->fragpa_lo = cpu_to_le32(frag_dmaaddr & 0xFFFFFFFF);
1906 rxd->fragpa_hi = cpu_to_le32(upper_32_bits(frag_dmaaddr));
1908 /* Any space left in the current big page for another frag? */
1909 if ((page_offset + rx_frag_size + rx_frag_size) >
1910 adapter->big_page_size) {
1912 page_info->last_frag = true;
1913 dma_unmap_addr_set(page_info, bus, page_dmaaddr);
1915 dma_unmap_addr_set(page_info, bus, frag_dmaaddr);
1918 prev_page_info = page_info;
1919 queue_head_inc(rxq);
1920 page_info = &rxo->page_info_tbl[rxq->head];
1923 /* Mark the last frag of a page when we break out of the above loop
1924 * with no more slots available in the RXQ
1927 prev_page_info->last_frag = true;
1928 dma_unmap_addr_set(prev_page_info, bus, page_dmaaddr);
1932 atomic_add(posted, &rxq->used);
1933 if (rxo->rx_post_starved)
1934 rxo->rx_post_starved = false;
1936 notify = min(256u, posted);
1937 be_rxq_notify(adapter, rxq->id, notify);
1940 } else if (atomic_read(&rxq->used) == 0) {
1941 /* Let be_worker replenish when memory is available */
1942 rxo->rx_post_starved = true;
1946 static struct be_eth_tx_compl *be_tx_compl_get(struct be_queue_info *tx_cq)
1948 struct be_eth_tx_compl *txcp = queue_tail_node(tx_cq);
1950 if (txcp->dw[offsetof(struct amap_eth_tx_compl, valid) / 32] == 0)
1954 be_dws_le_to_cpu(txcp, sizeof(*txcp));
1956 txcp->dw[offsetof(struct amap_eth_tx_compl, valid) / 32] = 0;
1958 queue_tail_inc(tx_cq);
1962 static u16 be_tx_compl_process(struct be_adapter *adapter,
1963 struct be_tx_obj *txo, u16 last_index)
1965 struct sk_buff **sent_skbs = txo->sent_skb_list;
1966 struct be_queue_info *txq = &txo->q;
1967 u16 frag_index, num_wrbs = 0;
1968 struct sk_buff *skb = NULL;
1969 bool unmap_skb_hdr = false;
1970 struct be_eth_wrb *wrb;
1973 if (sent_skbs[txq->tail]) {
1974 /* Free skb from prev req */
1976 dev_consume_skb_any(skb);
1977 skb = sent_skbs[txq->tail];
1978 sent_skbs[txq->tail] = NULL;
1979 queue_tail_inc(txq); /* skip hdr wrb */
1981 unmap_skb_hdr = true;
1983 wrb = queue_tail_node(txq);
1984 frag_index = txq->tail;
1985 unmap_tx_frag(&adapter->pdev->dev, wrb,
1986 (unmap_skb_hdr && skb_headlen(skb)));
1987 unmap_skb_hdr = false;
1988 queue_tail_inc(txq);
1990 } while (frag_index != last_index);
1991 dev_consume_skb_any(skb);
1996 /* Return the number of events in the event queue */
1997 static inline int events_get(struct be_eq_obj *eqo)
1999 struct be_eq_entry *eqe;
2003 eqe = queue_tail_node(&eqo->q);
2010 queue_tail_inc(&eqo->q);
2016 /* Leaves the EQ is disarmed state */
2017 static void be_eq_clean(struct be_eq_obj *eqo)
2019 int num = events_get(eqo);
2021 be_eq_notify(eqo->adapter, eqo->q.id, false, true, num);
2024 static void be_rx_cq_clean(struct be_rx_obj *rxo)
2026 struct be_rx_page_info *page_info;
2027 struct be_queue_info *rxq = &rxo->q;
2028 struct be_queue_info *rx_cq = &rxo->cq;
2029 struct be_rx_compl_info *rxcp;
2030 struct be_adapter *adapter = rxo->adapter;
2033 /* Consume pending rx completions.
2034 * Wait for the flush completion (identified by zero num_rcvd)
2035 * to arrive. Notify CQ even when there are no more CQ entries
2036 * for HW to flush partially coalesced CQ entries.
2037 * In Lancer, there is no need to wait for flush compl.
2040 rxcp = be_rx_compl_get(rxo);
2042 if (lancer_chip(adapter))
2045 if (flush_wait++ > 10 || be_hw_error(adapter)) {
2046 dev_warn(&adapter->pdev->dev,
2047 "did not receive flush compl\n");
2050 be_cq_notify(adapter, rx_cq->id, true, 0);
2053 be_rx_compl_discard(rxo, rxcp);
2054 be_cq_notify(adapter, rx_cq->id, false, 1);
2055 if (rxcp->num_rcvd == 0)
2060 /* After cleanup, leave the CQ in unarmed state */
2061 be_cq_notify(adapter, rx_cq->id, false, 0);
2063 /* Then free posted rx buffers that were not used */
2064 while (atomic_read(&rxq->used) > 0) {
2065 page_info = get_rx_page_info(rxo);
2066 put_page(page_info->page);
2067 memset(page_info, 0, sizeof(*page_info));
2069 BUG_ON(atomic_read(&rxq->used));
2074 static void be_tx_compl_clean(struct be_adapter *adapter)
2076 u16 end_idx, notified_idx, cmpl = 0, timeo = 0, num_wrbs = 0;
2077 struct device *dev = &adapter->pdev->dev;
2078 struct be_tx_obj *txo;
2079 struct be_queue_info *txq;
2080 struct be_eth_tx_compl *txcp;
2081 int i, pending_txqs;
2083 /* Stop polling for compls when HW has been silent for 10ms */
2085 pending_txqs = adapter->num_tx_qs;
2087 for_all_tx_queues(adapter, txo, i) {
2091 while ((txcp = be_tx_compl_get(&txo->cq))) {
2092 end_idx = GET_TX_COMPL_BITS(wrb_index, txcp);
2093 num_wrbs += be_tx_compl_process(adapter, txo,
2098 be_cq_notify(adapter, txo->cq.id, false, cmpl);
2099 atomic_sub(num_wrbs, &txq->used);
2102 if (atomic_read(&txq->used) == txo->pend_wrb_cnt)
2106 if (pending_txqs == 0 || ++timeo > 10 || be_hw_error(adapter))
2112 /* Free enqueued TX that was never notified to HW */
2113 for_all_tx_queues(adapter, txo, i) {
2116 if (atomic_read(&txq->used)) {
2117 dev_info(dev, "txq%d: cleaning %d pending tx-wrbs\n",
2118 i, atomic_read(&txq->used));
2119 notified_idx = txq->tail;
2120 end_idx = txq->tail;
2121 index_adv(&end_idx, atomic_read(&txq->used) - 1,
2123 /* Use the tx-compl process logic to handle requests
2124 * that were not sent to the HW.
2126 num_wrbs = be_tx_compl_process(adapter, txo, end_idx);
2127 atomic_sub(num_wrbs, &txq->used);
2128 BUG_ON(atomic_read(&txq->used));
2129 txo->pend_wrb_cnt = 0;
2130 /* Since hw was never notified of these requests,
2133 txq->head = notified_idx;
2134 txq->tail = notified_idx;
2139 static void be_evt_queues_destroy(struct be_adapter *adapter)
2141 struct be_eq_obj *eqo;
2144 for_all_evt_queues(adapter, eqo, i) {
2145 if (eqo->q.created) {
2147 be_cmd_q_destroy(adapter, &eqo->q, QTYPE_EQ);
2148 napi_hash_del(&eqo->napi);
2149 netif_napi_del(&eqo->napi);
2151 be_queue_free(adapter, &eqo->q);
2155 static int be_evt_queues_create(struct be_adapter *adapter)
2157 struct be_queue_info *eq;
2158 struct be_eq_obj *eqo;
2159 struct be_aic_obj *aic;
2162 adapter->num_evt_qs = min_t(u16, num_irqs(adapter),
2163 adapter->cfg_num_qs);
2165 for_all_evt_queues(adapter, eqo, i) {
2166 netif_napi_add(adapter->netdev, &eqo->napi, be_poll,
2168 napi_hash_add(&eqo->napi);
2169 aic = &adapter->aic_obj[i];
2170 eqo->adapter = adapter;
2172 aic->max_eqd = BE_MAX_EQD;
2176 rc = be_queue_alloc(adapter, eq, EVNT_Q_LEN,
2177 sizeof(struct be_eq_entry));
2181 rc = be_cmd_eq_create(adapter, eqo);
2188 static void be_mcc_queues_destroy(struct be_adapter *adapter)
2190 struct be_queue_info *q;
2192 q = &adapter->mcc_obj.q;
2194 be_cmd_q_destroy(adapter, q, QTYPE_MCCQ);
2195 be_queue_free(adapter, q);
2197 q = &adapter->mcc_obj.cq;
2199 be_cmd_q_destroy(adapter, q, QTYPE_CQ);
2200 be_queue_free(adapter, q);
2203 /* Must be called only after TX qs are created as MCC shares TX EQ */
2204 static int be_mcc_queues_create(struct be_adapter *adapter)
2206 struct be_queue_info *q, *cq;
2208 cq = &adapter->mcc_obj.cq;
2209 if (be_queue_alloc(adapter, cq, MCC_CQ_LEN,
2210 sizeof(struct be_mcc_compl)))
2213 /* Use the default EQ for MCC completions */
2214 if (be_cmd_cq_create(adapter, cq, &mcc_eqo(adapter)->q, true, 0))
2217 q = &adapter->mcc_obj.q;
2218 if (be_queue_alloc(adapter, q, MCC_Q_LEN, sizeof(struct be_mcc_wrb)))
2219 goto mcc_cq_destroy;
2221 if (be_cmd_mccq_create(adapter, q, cq))
2227 be_queue_free(adapter, q);
2229 be_cmd_q_destroy(adapter, cq, QTYPE_CQ);
2231 be_queue_free(adapter, cq);
2236 static void be_tx_queues_destroy(struct be_adapter *adapter)
2238 struct be_queue_info *q;
2239 struct be_tx_obj *txo;
2242 for_all_tx_queues(adapter, txo, i) {
2245 be_cmd_q_destroy(adapter, q, QTYPE_TXQ);
2246 be_queue_free(adapter, q);
2250 be_cmd_q_destroy(adapter, q, QTYPE_CQ);
2251 be_queue_free(adapter, q);
2255 static int be_tx_qs_create(struct be_adapter *adapter)
2257 struct be_queue_info *cq, *eq;
2258 struct be_tx_obj *txo;
2261 adapter->num_tx_qs = min(adapter->num_evt_qs, be_max_txqs(adapter));
2263 for_all_tx_queues(adapter, txo, i) {
2265 status = be_queue_alloc(adapter, cq, TX_CQ_LEN,
2266 sizeof(struct be_eth_tx_compl));
2270 u64_stats_init(&txo->stats.sync);
2271 u64_stats_init(&txo->stats.sync_compl);
2273 /* If num_evt_qs is less than num_tx_qs, then more than
2274 * one txq share an eq
2276 eq = &adapter->eq_obj[i % adapter->num_evt_qs].q;
2277 status = be_cmd_cq_create(adapter, cq, eq, false, 3);
2281 status = be_queue_alloc(adapter, &txo->q, TX_Q_LEN,
2282 sizeof(struct be_eth_wrb));
2286 status = be_cmd_txq_create(adapter, txo);
2291 dev_info(&adapter->pdev->dev, "created %d TX queue(s)\n",
2292 adapter->num_tx_qs);
2296 static void be_rx_cqs_destroy(struct be_adapter *adapter)
2298 struct be_queue_info *q;
2299 struct be_rx_obj *rxo;
2302 for_all_rx_queues(adapter, rxo, i) {
2305 be_cmd_q_destroy(adapter, q, QTYPE_CQ);
2306 be_queue_free(adapter, q);
2310 static int be_rx_cqs_create(struct be_adapter *adapter)
2312 struct be_queue_info *eq, *cq;
2313 struct be_rx_obj *rxo;
2316 /* We can create as many RSS rings as there are EQs. */
2317 adapter->num_rx_qs = adapter->num_evt_qs;
2319 /* We'll use RSS only if atleast 2 RSS rings are supported.
2320 * When RSS is used, we'll need a default RXQ for non-IP traffic.
2322 if (adapter->num_rx_qs > 1)
2323 adapter->num_rx_qs++;
2325 adapter->big_page_size = (1 << get_order(rx_frag_size)) * PAGE_SIZE;
2326 for_all_rx_queues(adapter, rxo, i) {
2327 rxo->adapter = adapter;
2329 rc = be_queue_alloc(adapter, cq, RX_CQ_LEN,
2330 sizeof(struct be_eth_rx_compl));
2334 u64_stats_init(&rxo->stats.sync);
2335 eq = &adapter->eq_obj[i % adapter->num_evt_qs].q;
2336 rc = be_cmd_cq_create(adapter, cq, eq, false, 3);
2341 dev_info(&adapter->pdev->dev,
2342 "created %d RSS queue(s) and 1 default RX queue\n",
2343 adapter->num_rx_qs - 1);
2347 static irqreturn_t be_intx(int irq, void *dev)
2349 struct be_eq_obj *eqo = dev;
2350 struct be_adapter *adapter = eqo->adapter;
2353 /* IRQ is not expected when NAPI is scheduled as the EQ
2354 * will not be armed.
2355 * But, this can happen on Lancer INTx where it takes
2356 * a while to de-assert INTx or in BE2 where occasionaly
2357 * an interrupt may be raised even when EQ is unarmed.
2358 * If NAPI is already scheduled, then counting & notifying
2359 * events will orphan them.
2361 if (napi_schedule_prep(&eqo->napi)) {
2362 num_evts = events_get(eqo);
2363 __napi_schedule(&eqo->napi);
2365 eqo->spurious_intr = 0;
2367 be_eq_notify(adapter, eqo->q.id, false, true, num_evts);
2369 /* Return IRQ_HANDLED only for the the first spurious intr
2370 * after a valid intr to stop the kernel from branding
2371 * this irq as a bad one!
2373 if (num_evts || eqo->spurious_intr++ == 0)
2379 static irqreturn_t be_msix(int irq, void *dev)
2381 struct be_eq_obj *eqo = dev;
2383 be_eq_notify(eqo->adapter, eqo->q.id, false, true, 0);
2384 napi_schedule(&eqo->napi);
2388 static inline bool do_gro(struct be_rx_compl_info *rxcp)
2390 return (rxcp->tcpf && !rxcp->err && rxcp->l4_csum) ? true : false;
2393 static int be_process_rx(struct be_rx_obj *rxo, struct napi_struct *napi,
2394 int budget, int polling)
2396 struct be_adapter *adapter = rxo->adapter;
2397 struct be_queue_info *rx_cq = &rxo->cq;
2398 struct be_rx_compl_info *rxcp;
2400 u32 frags_consumed = 0;
2402 for (work_done = 0; work_done < budget; work_done++) {
2403 rxcp = be_rx_compl_get(rxo);
2407 /* Is it a flush compl that has no data */
2408 if (unlikely(rxcp->num_rcvd == 0))
2411 /* Discard compl with partial DMA Lancer B0 */
2412 if (unlikely(!rxcp->pkt_size)) {
2413 be_rx_compl_discard(rxo, rxcp);
2417 /* On BE drop pkts that arrive due to imperfect filtering in
2418 * promiscuous mode on some skews
2420 if (unlikely(rxcp->port != adapter->port_num &&
2421 !lancer_chip(adapter))) {
2422 be_rx_compl_discard(rxo, rxcp);
2426 /* Don't do gro when we're busy_polling */
2427 if (do_gro(rxcp) && polling != BUSY_POLLING)
2428 be_rx_compl_process_gro(rxo, napi, rxcp);
2430 be_rx_compl_process(rxo, napi, rxcp);
2433 frags_consumed += rxcp->num_rcvd;
2434 be_rx_stats_update(rxo, rxcp);
2438 be_cq_notify(adapter, rx_cq->id, true, work_done);
2440 /* When an rx-obj gets into post_starved state, just
2441 * let be_worker do the posting.
2443 if (atomic_read(&rxo->q.used) < RX_FRAGS_REFILL_WM &&
2444 !rxo->rx_post_starved)
2445 be_post_rx_frags(rxo, GFP_ATOMIC,
2446 max_t(u32, MAX_RX_POST,
2453 static inline void be_update_tx_err(struct be_tx_obj *txo, u32 status)
2456 case BE_TX_COMP_HDR_PARSE_ERR:
2457 tx_stats(txo)->tx_hdr_parse_err++;
2459 case BE_TX_COMP_NDMA_ERR:
2460 tx_stats(txo)->tx_dma_err++;
2462 case BE_TX_COMP_ACL_ERR:
2463 tx_stats(txo)->tx_spoof_check_err++;
2468 static inline void lancer_update_tx_err(struct be_tx_obj *txo, u32 status)
2471 case LANCER_TX_COMP_LSO_ERR:
2472 tx_stats(txo)->tx_tso_err++;
2474 case LANCER_TX_COMP_HSW_DROP_MAC_ERR:
2475 case LANCER_TX_COMP_HSW_DROP_VLAN_ERR:
2476 tx_stats(txo)->tx_spoof_check_err++;
2478 case LANCER_TX_COMP_QINQ_ERR:
2479 tx_stats(txo)->tx_qinq_err++;
2481 case LANCER_TX_COMP_PARITY_ERR:
2482 tx_stats(txo)->tx_internal_parity_err++;
2484 case LANCER_TX_COMP_DMA_ERR:
2485 tx_stats(txo)->tx_dma_err++;
2490 static void be_process_tx(struct be_adapter *adapter, struct be_tx_obj *txo,
2493 struct be_eth_tx_compl *txcp;
2494 int num_wrbs = 0, work_done = 0;
2498 while ((txcp = be_tx_compl_get(&txo->cq))) {
2499 last_idx = GET_TX_COMPL_BITS(wrb_index, txcp);
2500 num_wrbs += be_tx_compl_process(adapter, txo, last_idx);
2503 compl_status = GET_TX_COMPL_BITS(status, txcp);
2505 if (lancer_chip(adapter))
2506 lancer_update_tx_err(txo, compl_status);
2508 be_update_tx_err(txo, compl_status);
2513 be_cq_notify(adapter, txo->cq.id, true, work_done);
2514 atomic_sub(num_wrbs, &txo->q.used);
2516 /* As Tx wrbs have been freed up, wake up netdev queue
2517 * if it was stopped due to lack of tx wrbs. */
2518 if (__netif_subqueue_stopped(adapter->netdev, idx) &&
2519 atomic_read(&txo->q.used) < txo->q.len / 2) {
2520 netif_wake_subqueue(adapter->netdev, idx);
2523 u64_stats_update_begin(&tx_stats(txo)->sync_compl);
2524 tx_stats(txo)->tx_compl += work_done;
2525 u64_stats_update_end(&tx_stats(txo)->sync_compl);
2529 int be_poll(struct napi_struct *napi, int budget)
2531 struct be_eq_obj *eqo = container_of(napi, struct be_eq_obj, napi);
2532 struct be_adapter *adapter = eqo->adapter;
2533 int max_work = 0, work, i, num_evts;
2534 struct be_rx_obj *rxo;
2535 struct be_tx_obj *txo;
2537 num_evts = events_get(eqo);
2539 for_all_tx_queues_on_eq(adapter, eqo, txo, i)
2540 be_process_tx(adapter, txo, i);
2542 if (be_lock_napi(eqo)) {
2543 /* This loop will iterate twice for EQ0 in which
2544 * completions of the last RXQ (default one) are also processed
2545 * For other EQs the loop iterates only once
2547 for_all_rx_queues_on_eq(adapter, eqo, rxo, i) {
2548 work = be_process_rx(rxo, napi, budget, NAPI_POLLING);
2549 max_work = max(work, max_work);
2551 be_unlock_napi(eqo);
2556 if (is_mcc_eqo(eqo))
2557 be_process_mcc(adapter);
2559 if (max_work < budget) {
2560 napi_complete(napi);
2561 be_eq_notify(adapter, eqo->q.id, true, false, num_evts);
2563 /* As we'll continue in polling mode, count and clear events */
2564 be_eq_notify(adapter, eqo->q.id, false, false, num_evts);
2569 #ifdef CONFIG_NET_RX_BUSY_POLL
2570 static int be_busy_poll(struct napi_struct *napi)
2572 struct be_eq_obj *eqo = container_of(napi, struct be_eq_obj, napi);
2573 struct be_adapter *adapter = eqo->adapter;
2574 struct be_rx_obj *rxo;
2577 if (!be_lock_busy_poll(eqo))
2578 return LL_FLUSH_BUSY;
2580 for_all_rx_queues_on_eq(adapter, eqo, rxo, i) {
2581 work = be_process_rx(rxo, napi, 4, BUSY_POLLING);
2586 be_unlock_busy_poll(eqo);
2591 void be_detect_error(struct be_adapter *adapter)
2593 u32 ue_lo = 0, ue_hi = 0, ue_lo_mask = 0, ue_hi_mask = 0;
2594 u32 sliport_status = 0, sliport_err1 = 0, sliport_err2 = 0;
2596 bool error_detected = false;
2597 struct device *dev = &adapter->pdev->dev;
2598 struct net_device *netdev = adapter->netdev;
2600 if (be_hw_error(adapter))
2603 if (lancer_chip(adapter)) {
2604 sliport_status = ioread32(adapter->db + SLIPORT_STATUS_OFFSET);
2605 if (sliport_status & SLIPORT_STATUS_ERR_MASK) {
2606 sliport_err1 = ioread32(adapter->db +
2607 SLIPORT_ERROR1_OFFSET);
2608 sliport_err2 = ioread32(adapter->db +
2609 SLIPORT_ERROR2_OFFSET);
2610 adapter->hw_error = true;
2611 /* Do not log error messages if its a FW reset */
2612 if (sliport_err1 == SLIPORT_ERROR_FW_RESET1 &&
2613 sliport_err2 == SLIPORT_ERROR_FW_RESET2) {
2614 dev_info(dev, "Firmware update in progress\n");
2616 error_detected = true;
2617 dev_err(dev, "Error detected in the card\n");
2618 dev_err(dev, "ERR: sliport status 0x%x\n",
2620 dev_err(dev, "ERR: sliport error1 0x%x\n",
2622 dev_err(dev, "ERR: sliport error2 0x%x\n",
2627 pci_read_config_dword(adapter->pdev,
2628 PCICFG_UE_STATUS_LOW, &ue_lo);
2629 pci_read_config_dword(adapter->pdev,
2630 PCICFG_UE_STATUS_HIGH, &ue_hi);
2631 pci_read_config_dword(adapter->pdev,
2632 PCICFG_UE_STATUS_LOW_MASK, &ue_lo_mask);
2633 pci_read_config_dword(adapter->pdev,
2634 PCICFG_UE_STATUS_HI_MASK, &ue_hi_mask);
2636 ue_lo = (ue_lo & ~ue_lo_mask);
2637 ue_hi = (ue_hi & ~ue_hi_mask);
2639 /* On certain platforms BE hardware can indicate spurious UEs.
2640 * Allow HW to stop working completely in case of a real UE.
2641 * Hence not setting the hw_error for UE detection.
2644 if (ue_lo || ue_hi) {
2645 error_detected = true;
2647 "Unrecoverable Error detected in the adapter");
2648 dev_err(dev, "Please reboot server to recover");
2649 if (skyhawk_chip(adapter))
2650 adapter->hw_error = true;
2651 for (i = 0; ue_lo; ue_lo >>= 1, i++) {
2653 dev_err(dev, "UE: %s bit set\n",
2654 ue_status_low_desc[i]);
2656 for (i = 0; ue_hi; ue_hi >>= 1, i++) {
2658 dev_err(dev, "UE: %s bit set\n",
2659 ue_status_hi_desc[i]);
2664 netif_carrier_off(netdev);
2667 static void be_msix_disable(struct be_adapter *adapter)
2669 if (msix_enabled(adapter)) {
2670 pci_disable_msix(adapter->pdev);
2671 adapter->num_msix_vec = 0;
2672 adapter->num_msix_roce_vec = 0;
2676 static int be_msix_enable(struct be_adapter *adapter)
2679 struct device *dev = &adapter->pdev->dev;
2681 /* If RoCE is supported, program the max number of NIC vectors that
2682 * may be configured via set-channels, along with vectors needed for
2683 * RoCe. Else, just program the number we'll use initially.
2685 if (be_roce_supported(adapter))
2686 num_vec = min_t(int, 2 * be_max_eqs(adapter),
2687 2 * num_online_cpus());
2689 num_vec = adapter->cfg_num_qs;
2691 for (i = 0; i < num_vec; i++)
2692 adapter->msix_entries[i].entry = i;
2694 num_vec = pci_enable_msix_range(adapter->pdev, adapter->msix_entries,
2695 MIN_MSIX_VECTORS, num_vec);
2699 if (be_roce_supported(adapter) && num_vec > MIN_MSIX_VECTORS) {
2700 adapter->num_msix_roce_vec = num_vec / 2;
2701 dev_info(dev, "enabled %d MSI-x vector(s) for RoCE\n",
2702 adapter->num_msix_roce_vec);
2705 adapter->num_msix_vec = num_vec - adapter->num_msix_roce_vec;
2707 dev_info(dev, "enabled %d MSI-x vector(s) for NIC\n",
2708 adapter->num_msix_vec);
2712 dev_warn(dev, "MSIx enable failed\n");
2714 /* INTx is not supported in VFs, so fail probe if enable_msix fails */
2715 if (!be_physfn(adapter))
2720 static inline int be_msix_vec_get(struct be_adapter *adapter,
2721 struct be_eq_obj *eqo)
2723 return adapter->msix_entries[eqo->msix_idx].vector;
2726 static int be_msix_register(struct be_adapter *adapter)
2728 struct net_device *netdev = adapter->netdev;
2729 struct be_eq_obj *eqo;
2732 for_all_evt_queues(adapter, eqo, i) {
2733 sprintf(eqo->desc, "%s-q%d", netdev->name, i);
2734 vec = be_msix_vec_get(adapter, eqo);
2735 status = request_irq(vec, be_msix, 0, eqo->desc, eqo);
2742 for (i--, eqo = &adapter->eq_obj[i]; i >= 0; i--, eqo--)
2743 free_irq(be_msix_vec_get(adapter, eqo), eqo);
2744 dev_warn(&adapter->pdev->dev, "MSIX Request IRQ failed - err %d\n",
2746 be_msix_disable(adapter);
2750 static int be_irq_register(struct be_adapter *adapter)
2752 struct net_device *netdev = adapter->netdev;
2755 if (msix_enabled(adapter)) {
2756 status = be_msix_register(adapter);
2759 /* INTx is not supported for VF */
2760 if (!be_physfn(adapter))
2764 /* INTx: only the first EQ is used */
2765 netdev->irq = adapter->pdev->irq;
2766 status = request_irq(netdev->irq, be_intx, IRQF_SHARED, netdev->name,
2767 &adapter->eq_obj[0]);
2769 dev_err(&adapter->pdev->dev,
2770 "INTx request IRQ failed - err %d\n", status);
2774 adapter->isr_registered = true;
2778 static void be_irq_unregister(struct be_adapter *adapter)
2780 struct net_device *netdev = adapter->netdev;
2781 struct be_eq_obj *eqo;
2784 if (!adapter->isr_registered)
2788 if (!msix_enabled(adapter)) {
2789 free_irq(netdev->irq, &adapter->eq_obj[0]);
2794 for_all_evt_queues(adapter, eqo, i)
2795 free_irq(be_msix_vec_get(adapter, eqo), eqo);
2798 adapter->isr_registered = false;
2801 static void be_rx_qs_destroy(struct be_adapter *adapter)
2803 struct be_queue_info *q;
2804 struct be_rx_obj *rxo;
2807 for_all_rx_queues(adapter, rxo, i) {
2810 be_cmd_rxq_destroy(adapter, q);
2811 be_rx_cq_clean(rxo);
2813 be_queue_free(adapter, q);
2817 static int be_close(struct net_device *netdev)
2819 struct be_adapter *adapter = netdev_priv(netdev);
2820 struct be_eq_obj *eqo;
2823 /* This protection is needed as be_close() may be called even when the
2824 * adapter is in cleared state (after eeh perm failure)
2826 if (!(adapter->flags & BE_FLAGS_SETUP_DONE))
2829 be_roce_dev_close(adapter);
2831 if (adapter->flags & BE_FLAGS_NAPI_ENABLED) {
2832 for_all_evt_queues(adapter, eqo, i) {
2833 napi_disable(&eqo->napi);
2834 be_disable_busy_poll(eqo);
2836 adapter->flags &= ~BE_FLAGS_NAPI_ENABLED;
2839 be_async_mcc_disable(adapter);
2841 /* Wait for all pending tx completions to arrive so that
2842 * all tx skbs are freed.
2844 netif_tx_disable(netdev);
2845 be_tx_compl_clean(adapter);
2847 be_rx_qs_destroy(adapter);
2849 for (i = 1; i < (adapter->uc_macs + 1); i++)
2850 be_cmd_pmac_del(adapter, adapter->if_handle,
2851 adapter->pmac_id[i], 0);
2852 adapter->uc_macs = 0;
2854 for_all_evt_queues(adapter, eqo, i) {
2855 if (msix_enabled(adapter))
2856 synchronize_irq(be_msix_vec_get(adapter, eqo));
2858 synchronize_irq(netdev->irq);
2862 be_irq_unregister(adapter);
2867 static int be_rx_qs_create(struct be_adapter *adapter)
2869 struct rss_info *rss = &adapter->rss_info;
2870 u8 rss_key[RSS_HASH_KEY_LEN];
2871 struct be_rx_obj *rxo;
2874 for_all_rx_queues(adapter, rxo, i) {
2875 rc = be_queue_alloc(adapter, &rxo->q, RX_Q_LEN,
2876 sizeof(struct be_eth_rx_d));
2881 /* The FW would like the default RXQ to be created first */
2882 rxo = default_rxo(adapter);
2883 rc = be_cmd_rxq_create(adapter, &rxo->q, rxo->cq.id, rx_frag_size,
2884 adapter->if_handle, false, &rxo->rss_id);
2888 for_all_rss_queues(adapter, rxo, i) {
2889 rc = be_cmd_rxq_create(adapter, &rxo->q, rxo->cq.id,
2890 rx_frag_size, adapter->if_handle,
2891 true, &rxo->rss_id);
2896 if (be_multi_rxq(adapter)) {
2897 for (j = 0; j < RSS_INDIR_TABLE_LEN;
2898 j += adapter->num_rx_qs - 1) {
2899 for_all_rss_queues(adapter, rxo, i) {
2900 if ((j + i) >= RSS_INDIR_TABLE_LEN)
2902 rss->rsstable[j + i] = rxo->rss_id;
2903 rss->rss_queue[j + i] = i;
2906 rss->rss_flags = RSS_ENABLE_TCP_IPV4 | RSS_ENABLE_IPV4 |
2907 RSS_ENABLE_TCP_IPV6 | RSS_ENABLE_IPV6;
2909 if (!BEx_chip(adapter))
2910 rss->rss_flags |= RSS_ENABLE_UDP_IPV4 |
2911 RSS_ENABLE_UDP_IPV6;
2913 /* Disable RSS, if only default RX Q is created */
2914 rss->rss_flags = RSS_ENABLE_NONE;
2917 netdev_rss_key_fill(rss_key, RSS_HASH_KEY_LEN);
2918 rc = be_cmd_rss_config(adapter, rss->rsstable, rss->rss_flags,
2921 rss->rss_flags = RSS_ENABLE_NONE;
2925 memcpy(rss->rss_hkey, rss_key, RSS_HASH_KEY_LEN);
2927 /* First time posting */
2928 for_all_rx_queues(adapter, rxo, i)
2929 be_post_rx_frags(rxo, GFP_KERNEL, MAX_RX_POST);
2933 static int be_open(struct net_device *netdev)
2935 struct be_adapter *adapter = netdev_priv(netdev);
2936 struct be_eq_obj *eqo;
2937 struct be_rx_obj *rxo;
2938 struct be_tx_obj *txo;
2942 status = be_rx_qs_create(adapter);
2946 status = be_irq_register(adapter);
2950 for_all_rx_queues(adapter, rxo, i)
2951 be_cq_notify(adapter, rxo->cq.id, true, 0);
2953 for_all_tx_queues(adapter, txo, i)
2954 be_cq_notify(adapter, txo->cq.id, true, 0);
2956 be_async_mcc_enable(adapter);
2958 for_all_evt_queues(adapter, eqo, i) {
2959 napi_enable(&eqo->napi);
2960 be_enable_busy_poll(eqo);
2961 be_eq_notify(adapter, eqo->q.id, true, true, 0);
2963 adapter->flags |= BE_FLAGS_NAPI_ENABLED;
2965 status = be_cmd_link_status_query(adapter, NULL, &link_status, 0);
2967 be_link_status_update(adapter, link_status);
2969 netif_tx_start_all_queues(netdev);
2970 be_roce_dev_open(adapter);
2972 #ifdef CONFIG_BE2NET_VXLAN
2973 if (skyhawk_chip(adapter))
2974 vxlan_get_rx_port(netdev);
2979 be_close(adapter->netdev);
2983 static int be_setup_wol(struct be_adapter *adapter, bool enable)
2985 struct be_dma_mem cmd;
2989 memset(mac, 0, ETH_ALEN);
2991 cmd.size = sizeof(struct be_cmd_req_acpi_wol_magic_config);
2992 cmd.va = dma_zalloc_coherent(&adapter->pdev->dev, cmd.size, &cmd.dma,
2998 status = pci_write_config_dword(adapter->pdev,
2999 PCICFG_PM_CONTROL_OFFSET,
3000 PCICFG_PM_CONTROL_MASK);
3002 dev_err(&adapter->pdev->dev,
3003 "Could not enable Wake-on-lan\n");
3004 dma_free_coherent(&adapter->pdev->dev, cmd.size, cmd.va,
3008 status = be_cmd_enable_magic_wol(adapter,
3009 adapter->netdev->dev_addr,
3011 pci_enable_wake(adapter->pdev, PCI_D3hot, 1);
3012 pci_enable_wake(adapter->pdev, PCI_D3cold, 1);
3014 status = be_cmd_enable_magic_wol(adapter, mac, &cmd);
3015 pci_enable_wake(adapter->pdev, PCI_D3hot, 0);
3016 pci_enable_wake(adapter->pdev, PCI_D3cold, 0);
3019 dma_free_coherent(&adapter->pdev->dev, cmd.size, cmd.va, cmd.dma);
3024 * Generate a seed MAC address from the PF MAC Address using jhash.
3025 * MAC Address for VFs are assigned incrementally starting from the seed.
3026 * These addresses are programmed in the ASIC by the PF and the VF driver
3027 * queries for the MAC address during its probe.
3029 static int be_vf_eth_addr_config(struct be_adapter *adapter)
3034 struct be_vf_cfg *vf_cfg;
3036 be_vf_eth_addr_generate(adapter, mac);
3038 for_all_vfs(adapter, vf_cfg, vf) {
3039 if (BEx_chip(adapter))
3040 status = be_cmd_pmac_add(adapter, mac,
3042 &vf_cfg->pmac_id, vf + 1);
3044 status = be_cmd_set_mac(adapter, mac, vf_cfg->if_handle,
3048 dev_err(&adapter->pdev->dev,
3049 "Mac address assignment failed for VF %d\n",
3052 memcpy(vf_cfg->mac_addr, mac, ETH_ALEN);
3059 static int be_vfs_mac_query(struct be_adapter *adapter)
3063 struct be_vf_cfg *vf_cfg;
3065 for_all_vfs(adapter, vf_cfg, vf) {
3066 status = be_cmd_get_active_mac(adapter, vf_cfg->pmac_id,
3067 mac, vf_cfg->if_handle,
3071 memcpy(vf_cfg->mac_addr, mac, ETH_ALEN);
3076 static void be_vf_clear(struct be_adapter *adapter)
3078 struct be_vf_cfg *vf_cfg;
3081 if (pci_vfs_assigned(adapter->pdev)) {
3082 dev_warn(&adapter->pdev->dev,
3083 "VFs are assigned to VMs: not disabling VFs\n");
3087 pci_disable_sriov(adapter->pdev);
3089 for_all_vfs(adapter, vf_cfg, vf) {
3090 if (BEx_chip(adapter))
3091 be_cmd_pmac_del(adapter, vf_cfg->if_handle,
3092 vf_cfg->pmac_id, vf + 1);
3094 be_cmd_set_mac(adapter, NULL, vf_cfg->if_handle,
3097 be_cmd_if_destroy(adapter, vf_cfg->if_handle, vf + 1);
3100 kfree(adapter->vf_cfg);
3101 adapter->num_vfs = 0;
3102 adapter->flags &= ~BE_FLAGS_SRIOV_ENABLED;
3105 static void be_clear_queues(struct be_adapter *adapter)
3107 be_mcc_queues_destroy(adapter);
3108 be_rx_cqs_destroy(adapter);
3109 be_tx_queues_destroy(adapter);
3110 be_evt_queues_destroy(adapter);
3113 static void be_cancel_worker(struct be_adapter *adapter)
3115 if (adapter->flags & BE_FLAGS_WORKER_SCHEDULED) {
3116 cancel_delayed_work_sync(&adapter->work);
3117 adapter->flags &= ~BE_FLAGS_WORKER_SCHEDULED;
3121 static void be_mac_clear(struct be_adapter *adapter)
3125 if (adapter->pmac_id) {
3126 for (i = 0; i < (adapter->uc_macs + 1); i++)
3127 be_cmd_pmac_del(adapter, adapter->if_handle,
3128 adapter->pmac_id[i], 0);
3129 adapter->uc_macs = 0;
3131 kfree(adapter->pmac_id);
3132 adapter->pmac_id = NULL;
3136 #ifdef CONFIG_BE2NET_VXLAN
3137 static void be_disable_vxlan_offloads(struct be_adapter *adapter)
3139 struct net_device *netdev = adapter->netdev;
3141 if (adapter->flags & BE_FLAGS_VXLAN_OFFLOADS)
3142 be_cmd_manage_iface(adapter, adapter->if_handle,
3143 OP_CONVERT_TUNNEL_TO_NORMAL);
3145 if (adapter->vxlan_port)
3146 be_cmd_set_vxlan_port(adapter, 0);
3148 adapter->flags &= ~BE_FLAGS_VXLAN_OFFLOADS;
3149 adapter->vxlan_port = 0;
3151 netdev->hw_enc_features = 0;
3152 netdev->hw_features &= ~(NETIF_F_GSO_UDP_TUNNEL);
3153 netdev->features &= ~(NETIF_F_GSO_UDP_TUNNEL);
3157 static int be_clear(struct be_adapter *adapter)
3159 be_cancel_worker(adapter);
3161 if (sriov_enabled(adapter))
3162 be_vf_clear(adapter);
3164 /* Re-configure FW to distribute resources evenly across max-supported
3165 * number of VFs, only when VFs are not already enabled.
3167 if (be_physfn(adapter) && !pci_vfs_assigned(adapter->pdev))
3168 be_cmd_set_sriov_config(adapter, adapter->pool_res,
3169 pci_sriov_get_totalvfs(adapter->pdev));
3171 #ifdef CONFIG_BE2NET_VXLAN
3172 be_disable_vxlan_offloads(adapter);
3174 /* delete the primary mac along with the uc-mac list */
3175 be_mac_clear(adapter);
3177 be_cmd_if_destroy(adapter, adapter->if_handle, 0);
3179 be_clear_queues(adapter);
3181 be_msix_disable(adapter);
3182 adapter->flags &= ~BE_FLAGS_SETUP_DONE;
3186 static int be_if_create(struct be_adapter *adapter, u32 *if_handle,
3187 u32 cap_flags, u32 vf)
3192 en_flags = BE_IF_FLAGS_UNTAGGED | BE_IF_FLAGS_BROADCAST |
3193 BE_IF_FLAGS_MULTICAST | BE_IF_FLAGS_PASS_L3L4_ERRORS |
3196 en_flags &= cap_flags;
3198 status = be_cmd_if_create(adapter, cap_flags, en_flags,
3204 static int be_vfs_if_create(struct be_adapter *adapter)
3206 struct be_resources res = {0};
3207 struct be_vf_cfg *vf_cfg;
3211 /* If a FW profile exists, then cap_flags are updated */
3212 cap_flags = BE_IF_FLAGS_UNTAGGED | BE_IF_FLAGS_BROADCAST |
3213 BE_IF_FLAGS_MULTICAST;
3215 for_all_vfs(adapter, vf_cfg, vf) {
3216 if (!BE3_chip(adapter)) {
3217 status = be_cmd_get_profile_config(adapter, &res,
3220 cap_flags = res.if_cap_flags;
3223 status = be_if_create(adapter, &vf_cfg->if_handle,
3232 static int be_vf_setup_init(struct be_adapter *adapter)
3234 struct be_vf_cfg *vf_cfg;
3237 adapter->vf_cfg = kcalloc(adapter->num_vfs, sizeof(*vf_cfg),
3239 if (!adapter->vf_cfg)
3242 for_all_vfs(adapter, vf_cfg, vf) {
3243 vf_cfg->if_handle = -1;
3244 vf_cfg->pmac_id = -1;
3249 static int be_vf_setup(struct be_adapter *adapter)
3251 struct device *dev = &adapter->pdev->dev;
3252 struct be_vf_cfg *vf_cfg;
3253 int status, old_vfs, vf;
3256 old_vfs = pci_num_vf(adapter->pdev);
3258 status = be_vf_setup_init(adapter);
3263 for_all_vfs(adapter, vf_cfg, vf) {
3264 status = be_cmd_get_if_id(adapter, vf_cfg, vf);
3269 status = be_vfs_mac_query(adapter);
3273 status = be_vfs_if_create(adapter);
3277 status = be_vf_eth_addr_config(adapter);
3282 for_all_vfs(adapter, vf_cfg, vf) {
3283 /* Allow VFs to programs MAC/VLAN filters */
3284 status = be_cmd_get_fn_privileges(adapter, &privileges, vf + 1);
3285 if (!status && !(privileges & BE_PRIV_FILTMGMT)) {
3286 status = be_cmd_set_fn_privileges(adapter,
3291 dev_info(dev, "VF%d has FILTMGMT privilege\n",
3295 /* Allow full available bandwidth */
3297 be_cmd_config_qos(adapter, 0, 0, vf + 1);
3300 be_cmd_enable_vf(adapter, vf + 1);
3301 be_cmd_set_logical_link_config(adapter,
3302 IFLA_VF_LINK_STATE_AUTO,
3308 status = pci_enable_sriov(adapter->pdev, adapter->num_vfs);
3310 dev_err(dev, "SRIOV enable failed\n");
3311 adapter->num_vfs = 0;
3316 adapter->flags |= BE_FLAGS_SRIOV_ENABLED;
3319 dev_err(dev, "VF setup failed\n");
3320 be_vf_clear(adapter);
3324 /* Converting function_mode bits on BE3 to SH mc_type enums */
3326 static u8 be_convert_mc_type(u32 function_mode)
3328 if (function_mode & VNIC_MODE && function_mode & QNQ_MODE)
3330 else if (function_mode & QNQ_MODE)
3332 else if (function_mode & VNIC_MODE)
3334 else if (function_mode & UMC_ENABLED)
3340 /* On BE2/BE3 FW does not suggest the supported limits */
3341 static void BEx_get_resources(struct be_adapter *adapter,
3342 struct be_resources *res)
3344 bool use_sriov = adapter->num_vfs ? 1 : 0;
3346 if (be_physfn(adapter))
3347 res->max_uc_mac = BE_UC_PMAC_COUNT;
3349 res->max_uc_mac = BE_VF_UC_PMAC_COUNT;
3351 adapter->mc_type = be_convert_mc_type(adapter->function_mode);
3353 if (be_is_mc(adapter)) {
3354 /* Assuming that there are 4 channels per port,
3355 * when multi-channel is enabled
3357 if (be_is_qnq_mode(adapter))
3358 res->max_vlans = BE_NUM_VLANS_SUPPORTED/8;
3360 /* In a non-qnq multichannel mode, the pvid
3361 * takes up one vlan entry
3363 res->max_vlans = (BE_NUM_VLANS_SUPPORTED / 4) - 1;
3365 res->max_vlans = BE_NUM_VLANS_SUPPORTED;
3368 res->max_mcast_mac = BE_MAX_MC;
3370 /* 1) For BE3 1Gb ports, FW does not support multiple TXQs
3371 * 2) Create multiple TX rings on a BE3-R multi-channel interface
3372 * *only* if it is RSS-capable.
3374 if (BE2_chip(adapter) || use_sriov || (adapter->port_num > 1) ||
3375 !be_physfn(adapter) || (be_is_mc(adapter) &&
3376 !(adapter->function_caps & BE_FUNCTION_CAPS_RSS))) {
3378 } else if (adapter->function_caps & BE_FUNCTION_CAPS_SUPER_NIC) {
3379 struct be_resources super_nic_res = {0};
3381 /* On a SuperNIC profile, the driver needs to use the
3382 * GET_PROFILE_CONFIG cmd to query the per-function TXQ limits
3384 be_cmd_get_profile_config(adapter, &super_nic_res, 0);
3385 /* Some old versions of BE3 FW don't report max_tx_qs value */
3386 res->max_tx_qs = super_nic_res.max_tx_qs ? : BE3_MAX_TX_QS;
3388 res->max_tx_qs = BE3_MAX_TX_QS;
3391 if ((adapter->function_caps & BE_FUNCTION_CAPS_RSS) &&
3392 !use_sriov && be_physfn(adapter))
3393 res->max_rss_qs = (adapter->be3_native) ?
3394 BE3_MAX_RSS_QS : BE2_MAX_RSS_QS;
3395 res->max_rx_qs = res->max_rss_qs + 1;
3397 if (be_physfn(adapter))
3398 res->max_evt_qs = (be_max_vfs(adapter) > 0) ?
3399 BE3_SRIOV_MAX_EVT_QS : BE3_MAX_EVT_QS;
3401 res->max_evt_qs = 1;
3403 res->if_cap_flags = BE_IF_CAP_FLAGS_WANT;
3404 if (!(adapter->function_caps & BE_FUNCTION_CAPS_RSS))
3405 res->if_cap_flags &= ~BE_IF_FLAGS_RSS;
3408 static void be_setup_init(struct be_adapter *adapter)
3410 adapter->vlan_prio_bmap = 0xff;
3411 adapter->phy.link_speed = -1;
3412 adapter->if_handle = -1;
3413 adapter->be3_native = false;
3414 adapter->promiscuous = false;
3415 if (be_physfn(adapter))
3416 adapter->cmd_privileges = MAX_PRIVILEGES;
3418 adapter->cmd_privileges = MIN_PRIVILEGES;
3421 static int be_get_sriov_config(struct be_adapter *adapter)
3423 struct device *dev = &adapter->pdev->dev;
3424 struct be_resources res = {0};
3425 int max_vfs, old_vfs;
3427 /* Some old versions of BE3 FW don't report max_vfs value */
3428 be_cmd_get_profile_config(adapter, &res, 0);
3430 if (BE3_chip(adapter) && !res.max_vfs) {
3431 max_vfs = pci_sriov_get_totalvfs(adapter->pdev);
3432 res.max_vfs = max_vfs > 0 ? min(MAX_VFS, max_vfs) : 0;
3435 adapter->pool_res = res;
3437 if (!be_max_vfs(adapter)) {
3439 dev_warn(dev, "SRIOV is disabled. Ignoring num_vfs\n");
3440 adapter->num_vfs = 0;
3444 pci_sriov_set_totalvfs(adapter->pdev, be_max_vfs(adapter));
3446 /* validate num_vfs module param */
3447 old_vfs = pci_num_vf(adapter->pdev);
3449 dev_info(dev, "%d VFs are already enabled\n", old_vfs);
3450 if (old_vfs != num_vfs)
3451 dev_warn(dev, "Ignoring num_vfs=%d setting\n", num_vfs);
3452 adapter->num_vfs = old_vfs;
3454 if (num_vfs > be_max_vfs(adapter)) {
3455 dev_info(dev, "Resources unavailable to init %d VFs\n",
3457 dev_info(dev, "Limiting to %d VFs\n",
3458 be_max_vfs(adapter));
3460 adapter->num_vfs = min_t(u16, num_vfs, be_max_vfs(adapter));
3466 static int be_get_resources(struct be_adapter *adapter)
3468 struct device *dev = &adapter->pdev->dev;
3469 struct be_resources res = {0};
3472 if (BEx_chip(adapter)) {
3473 BEx_get_resources(adapter, &res);
3477 /* For Lancer, SH etc read per-function resource limits from FW.
3478 * GET_FUNC_CONFIG returns per function guaranteed limits.
3479 * GET_PROFILE_CONFIG returns PCI-E related limits PF-pool limits
3481 if (!BEx_chip(adapter)) {
3482 status = be_cmd_get_func_config(adapter, &res);
3486 /* If RoCE may be enabled stash away half the EQs for RoCE */
3487 if (be_roce_supported(adapter))
3488 res.max_evt_qs /= 2;
3492 dev_info(dev, "Max: txqs %d, rxqs %d, rss %d, eqs %d, vfs %d\n",
3493 be_max_txqs(adapter), be_max_rxqs(adapter),
3494 be_max_rss(adapter), be_max_eqs(adapter),
3495 be_max_vfs(adapter));
3496 dev_info(dev, "Max: uc-macs %d, mc-macs %d, vlans %d\n",
3497 be_max_uc(adapter), be_max_mc(adapter),
3498 be_max_vlans(adapter));
3503 static void be_sriov_config(struct be_adapter *adapter)
3505 struct device *dev = &adapter->pdev->dev;
3508 status = be_get_sriov_config(adapter);
3510 dev_err(dev, "Failed to query SR-IOV configuration\n");
3511 dev_err(dev, "SR-IOV cannot be enabled\n");
3515 /* When the HW is in SRIOV capable configuration, the PF-pool
3516 * resources are equally distributed across the max-number of
3517 * VFs. The user may request only a subset of the max-vfs to be
3518 * enabled. Based on num_vfs, redistribute the resources across
3519 * num_vfs so that each VF will have access to more number of
3520 * resources. This facility is not available in BE3 FW.
3521 * Also, this is done by FW in Lancer chip.
3523 if (be_max_vfs(adapter) && !pci_num_vf(adapter->pdev)) {
3524 status = be_cmd_set_sriov_config(adapter,
3528 dev_err(dev, "Failed to optimize SR-IOV resources\n");
3532 static int be_get_config(struct be_adapter *adapter)
3537 status = be_cmd_query_fw_cfg(adapter);
3541 if (be_physfn(adapter)) {
3542 status = be_cmd_get_active_profile(adapter, &profile_id);
3544 dev_info(&adapter->pdev->dev,
3545 "Using profile 0x%x\n", profile_id);
3548 if (!BE2_chip(adapter) && be_physfn(adapter))
3549 be_sriov_config(adapter);
3551 status = be_get_resources(adapter);
3555 adapter->pmac_id = kcalloc(be_max_uc(adapter),
3556 sizeof(*adapter->pmac_id), GFP_KERNEL);
3557 if (!adapter->pmac_id)
3560 /* Sanitize cfg_num_qs based on HW and platform limits */
3561 adapter->cfg_num_qs = min(adapter->cfg_num_qs, be_max_qs(adapter));
3566 static int be_mac_setup(struct be_adapter *adapter)
3571 if (is_zero_ether_addr(adapter->netdev->dev_addr)) {
3572 status = be_cmd_get_perm_mac(adapter, mac);
3576 memcpy(adapter->netdev->dev_addr, mac, ETH_ALEN);
3577 memcpy(adapter->netdev->perm_addr, mac, ETH_ALEN);
3579 /* Maybe the HW was reset; dev_addr must be re-programmed */
3580 memcpy(mac, adapter->netdev->dev_addr, ETH_ALEN);
3583 /* For BE3-R VFs, the PF programs the initial MAC address */
3584 if (!(BEx_chip(adapter) && be_virtfn(adapter)))
3585 be_cmd_pmac_add(adapter, mac, adapter->if_handle,
3586 &adapter->pmac_id[0], 0);
3590 static void be_schedule_worker(struct be_adapter *adapter)
3592 schedule_delayed_work(&adapter->work, msecs_to_jiffies(1000));
3593 adapter->flags |= BE_FLAGS_WORKER_SCHEDULED;
3596 static int be_setup_queues(struct be_adapter *adapter)
3598 struct net_device *netdev = adapter->netdev;
3601 status = be_evt_queues_create(adapter);
3605 status = be_tx_qs_create(adapter);
3609 status = be_rx_cqs_create(adapter);
3613 status = be_mcc_queues_create(adapter);
3617 status = netif_set_real_num_rx_queues(netdev, adapter->num_rx_qs);
3621 status = netif_set_real_num_tx_queues(netdev, adapter->num_tx_qs);
3627 dev_err(&adapter->pdev->dev, "queue_setup failed\n");
3631 int be_update_queues(struct be_adapter *adapter)
3633 struct net_device *netdev = adapter->netdev;
3636 if (netif_running(netdev))
3639 be_cancel_worker(adapter);
3641 /* If any vectors have been shared with RoCE we cannot re-program
3644 if (!adapter->num_msix_roce_vec)
3645 be_msix_disable(adapter);
3647 be_clear_queues(adapter);
3649 if (!msix_enabled(adapter)) {
3650 status = be_msix_enable(adapter);
3655 status = be_setup_queues(adapter);
3659 be_schedule_worker(adapter);
3661 if (netif_running(netdev))
3662 status = be_open(netdev);
3667 static int be_setup(struct be_adapter *adapter)
3669 struct device *dev = &adapter->pdev->dev;
3672 be_setup_init(adapter);
3674 if (!lancer_chip(adapter))
3675 be_cmd_req_native_mode(adapter);
3677 status = be_get_config(adapter);
3681 status = be_msix_enable(adapter);
3685 status = be_if_create(adapter, &adapter->if_handle,
3686 be_if_cap_flags(adapter), 0);
3690 /* Updating real_num_tx/rx_queues() requires rtnl_lock() */
3692 status = be_setup_queues(adapter);
3697 be_cmd_get_fn_privileges(adapter, &adapter->cmd_privileges, 0);
3699 status = be_mac_setup(adapter);
3703 be_cmd_get_fw_ver(adapter);
3704 dev_info(dev, "FW version is %s\n", adapter->fw_ver);
3706 if (BE2_chip(adapter) && fw_major_num(adapter->fw_ver) < 4) {
3707 dev_err(dev, "Firmware on card is old(%s), IRQs may not work",
3709 dev_err(dev, "Please upgrade firmware to version >= 4.0\n");
3712 if (adapter->vlans_added)
3713 be_vid_config(adapter);
3715 be_set_rx_mode(adapter->netdev);
3717 be_cmd_get_acpi_wol_cap(adapter);
3719 status = be_cmd_set_flow_control(adapter, adapter->tx_fc,
3722 be_cmd_get_flow_control(adapter, &adapter->tx_fc,
3725 dev_info(&adapter->pdev->dev, "HW Flow control - TX:%d RX:%d\n",
3726 adapter->tx_fc, adapter->rx_fc);
3728 if (be_physfn(adapter))
3729 be_cmd_set_logical_link_config(adapter,
3730 IFLA_VF_LINK_STATE_AUTO, 0);
3732 if (adapter->num_vfs)
3733 be_vf_setup(adapter);
3735 status = be_cmd_get_phy_info(adapter);
3736 if (!status && be_pause_supported(adapter))
3737 adapter->phy.fc_autoneg = 1;
3739 be_schedule_worker(adapter);
3740 adapter->flags |= BE_FLAGS_SETUP_DONE;
3747 #ifdef CONFIG_NET_POLL_CONTROLLER
3748 static void be_netpoll(struct net_device *netdev)
3750 struct be_adapter *adapter = netdev_priv(netdev);
3751 struct be_eq_obj *eqo;
3754 for_all_evt_queues(adapter, eqo, i) {
3755 be_eq_notify(eqo->adapter, eqo->q.id, false, true, 0);
3756 napi_schedule(&eqo->napi);
3761 static char flash_cookie[2][16] = {"*** SE FLAS", "H DIRECTORY *** "};
3763 static bool phy_flashing_required(struct be_adapter *adapter)
3765 return (adapter->phy.phy_type == PHY_TYPE_TN_8022 &&
3766 adapter->phy.interface_type == PHY_TYPE_BASET_10GB);
3769 static bool is_comp_in_ufi(struct be_adapter *adapter,
3770 struct flash_section_info *fsec, int type)
3772 int i = 0, img_type = 0;
3773 struct flash_section_info_g2 *fsec_g2 = NULL;
3775 if (BE2_chip(adapter))
3776 fsec_g2 = (struct flash_section_info_g2 *)fsec;
3778 for (i = 0; i < MAX_FLASH_COMP; i++) {
3780 img_type = le32_to_cpu(fsec_g2->fsec_entry[i].type);
3782 img_type = le32_to_cpu(fsec->fsec_entry[i].type);
3784 if (img_type == type)
3791 static struct flash_section_info *get_fsec_info(struct be_adapter *adapter,
3793 const struct firmware *fw)
3795 struct flash_section_info *fsec = NULL;
3796 const u8 *p = fw->data;
3799 while (p < (fw->data + fw->size)) {
3800 fsec = (struct flash_section_info *)p;
3801 if (!memcmp(flash_cookie, fsec->cookie, sizeof(flash_cookie)))
3808 static int be_check_flash_crc(struct be_adapter *adapter, const u8 *p,
3809 u32 img_offset, u32 img_size, int hdr_size,
3810 u16 img_optype, bool *crc_match)
3816 status = be_cmd_get_flash_crc(adapter, crc, img_optype, img_size - 4);
3820 crc_offset = hdr_size + img_offset + img_size - 4;
3822 /* Skip flashing, if crc of flashed region matches */
3823 if (!memcmp(crc, p + crc_offset, 4))
3831 static int be_flash(struct be_adapter *adapter, const u8 *img,
3832 struct be_dma_mem *flash_cmd, int optype, int img_size)
3834 struct be_cmd_write_flashrom *req = flash_cmd->va;
3835 u32 total_bytes, flash_op, num_bytes;
3838 total_bytes = img_size;
3839 while (total_bytes) {
3840 num_bytes = min_t(u32, 32*1024, total_bytes);
3842 total_bytes -= num_bytes;
3845 if (optype == OPTYPE_PHY_FW)
3846 flash_op = FLASHROM_OPER_PHY_FLASH;
3848 flash_op = FLASHROM_OPER_FLASH;
3850 if (optype == OPTYPE_PHY_FW)
3851 flash_op = FLASHROM_OPER_PHY_SAVE;
3853 flash_op = FLASHROM_OPER_SAVE;
3856 memcpy(req->data_buf, img, num_bytes);
3858 status = be_cmd_write_flashrom(adapter, flash_cmd, optype,
3859 flash_op, num_bytes);
3860 if (base_status(status) == MCC_STATUS_ILLEGAL_REQUEST &&
3861 optype == OPTYPE_PHY_FW)
3869 /* For BE2, BE3 and BE3-R */
3870 static int be_flash_BEx(struct be_adapter *adapter,
3871 const struct firmware *fw,
3872 struct be_dma_mem *flash_cmd, int num_of_images)
3874 int img_hdrs_size = (num_of_images * sizeof(struct image_hdr));
3875 struct device *dev = &adapter->pdev->dev;
3876 struct flash_section_info *fsec = NULL;
3877 int status, i, filehdr_size, num_comp;
3878 const struct flash_comp *pflashcomp;
3882 struct flash_comp gen3_flash_types[] = {
3883 { FLASH_iSCSI_PRIMARY_IMAGE_START_g3, OPTYPE_ISCSI_ACTIVE,
3884 FLASH_IMAGE_MAX_SIZE_g3, IMAGE_FIRMWARE_iSCSI},
3885 { FLASH_REDBOOT_START_g3, OPTYPE_REDBOOT,
3886 FLASH_REDBOOT_IMAGE_MAX_SIZE_g3, IMAGE_BOOT_CODE},
3887 { FLASH_iSCSI_BIOS_START_g3, OPTYPE_BIOS,
3888 FLASH_BIOS_IMAGE_MAX_SIZE_g3, IMAGE_OPTION_ROM_ISCSI},
3889 { FLASH_PXE_BIOS_START_g3, OPTYPE_PXE_BIOS,
3890 FLASH_BIOS_IMAGE_MAX_SIZE_g3, IMAGE_OPTION_ROM_PXE},
3891 { FLASH_FCoE_BIOS_START_g3, OPTYPE_FCOE_BIOS,
3892 FLASH_BIOS_IMAGE_MAX_SIZE_g3, IMAGE_OPTION_ROM_FCoE},
3893 { FLASH_iSCSI_BACKUP_IMAGE_START_g3, OPTYPE_ISCSI_BACKUP,
3894 FLASH_IMAGE_MAX_SIZE_g3, IMAGE_FIRMWARE_BACKUP_iSCSI},
3895 { FLASH_FCoE_PRIMARY_IMAGE_START_g3, OPTYPE_FCOE_FW_ACTIVE,
3896 FLASH_IMAGE_MAX_SIZE_g3, IMAGE_FIRMWARE_FCoE},
3897 { FLASH_FCoE_BACKUP_IMAGE_START_g3, OPTYPE_FCOE_FW_BACKUP,
3898 FLASH_IMAGE_MAX_SIZE_g3, IMAGE_FIRMWARE_BACKUP_FCoE},
3899 { FLASH_NCSI_START_g3, OPTYPE_NCSI_FW,
3900 FLASH_NCSI_IMAGE_MAX_SIZE_g3, IMAGE_NCSI},
3901 { FLASH_PHY_FW_START_g3, OPTYPE_PHY_FW,
3902 FLASH_PHY_FW_IMAGE_MAX_SIZE_g3, IMAGE_FIRMWARE_PHY}
3905 struct flash_comp gen2_flash_types[] = {
3906 { FLASH_iSCSI_PRIMARY_IMAGE_START_g2, OPTYPE_ISCSI_ACTIVE,
3907 FLASH_IMAGE_MAX_SIZE_g2, IMAGE_FIRMWARE_iSCSI},
3908 { FLASH_REDBOOT_START_g2, OPTYPE_REDBOOT,
3909 FLASH_REDBOOT_IMAGE_MAX_SIZE_g2, IMAGE_BOOT_CODE},
3910 { FLASH_iSCSI_BIOS_START_g2, OPTYPE_BIOS,
3911 FLASH_BIOS_IMAGE_MAX_SIZE_g2, IMAGE_OPTION_ROM_ISCSI},
3912 { FLASH_PXE_BIOS_START_g2, OPTYPE_PXE_BIOS,
3913 FLASH_BIOS_IMAGE_MAX_SIZE_g2, IMAGE_OPTION_ROM_PXE},
3914 { FLASH_FCoE_BIOS_START_g2, OPTYPE_FCOE_BIOS,
3915 FLASH_BIOS_IMAGE_MAX_SIZE_g2, IMAGE_OPTION_ROM_FCoE},
3916 { FLASH_iSCSI_BACKUP_IMAGE_START_g2, OPTYPE_ISCSI_BACKUP,
3917 FLASH_IMAGE_MAX_SIZE_g2, IMAGE_FIRMWARE_BACKUP_iSCSI},
3918 { FLASH_FCoE_PRIMARY_IMAGE_START_g2, OPTYPE_FCOE_FW_ACTIVE,
3919 FLASH_IMAGE_MAX_SIZE_g2, IMAGE_FIRMWARE_FCoE},
3920 { FLASH_FCoE_BACKUP_IMAGE_START_g2, OPTYPE_FCOE_FW_BACKUP,
3921 FLASH_IMAGE_MAX_SIZE_g2, IMAGE_FIRMWARE_BACKUP_FCoE}
3924 if (BE3_chip(adapter)) {
3925 pflashcomp = gen3_flash_types;
3926 filehdr_size = sizeof(struct flash_file_hdr_g3);
3927 num_comp = ARRAY_SIZE(gen3_flash_types);
3929 pflashcomp = gen2_flash_types;
3930 filehdr_size = sizeof(struct flash_file_hdr_g2);
3931 num_comp = ARRAY_SIZE(gen2_flash_types);
3934 /* Get flash section info*/
3935 fsec = get_fsec_info(adapter, filehdr_size + img_hdrs_size, fw);
3937 dev_err(dev, "Invalid Cookie. FW image may be corrupted\n");
3940 for (i = 0; i < num_comp; i++) {
3941 if (!is_comp_in_ufi(adapter, fsec, pflashcomp[i].img_type))
3944 if ((pflashcomp[i].optype == OPTYPE_NCSI_FW) &&
3945 memcmp(adapter->fw_ver, "3.102.148.0", 11) < 0)
3948 if (pflashcomp[i].optype == OPTYPE_PHY_FW &&
3949 !phy_flashing_required(adapter))
3952 if (pflashcomp[i].optype == OPTYPE_REDBOOT) {
3953 status = be_check_flash_crc(adapter, fw->data,
3954 pflashcomp[i].offset,
3958 OPTYPE_REDBOOT, &crc_match);
3961 "Could not get CRC for 0x%x region\n",
3962 pflashcomp[i].optype);
3970 p = fw->data + filehdr_size + pflashcomp[i].offset +
3972 if (p + pflashcomp[i].size > fw->data + fw->size)
3975 status = be_flash(adapter, p, flash_cmd, pflashcomp[i].optype,
3976 pflashcomp[i].size);
3978 dev_err(dev, "Flashing section type 0x%x failed\n",
3979 pflashcomp[i].img_type);
3986 static u16 be_get_img_optype(struct flash_section_entry fsec_entry)
3988 u32 img_type = le32_to_cpu(fsec_entry.type);
3989 u16 img_optype = le16_to_cpu(fsec_entry.optype);
3991 if (img_optype != 0xFFFF)
3995 case IMAGE_FIRMWARE_iSCSI:
3996 img_optype = OPTYPE_ISCSI_ACTIVE;
3998 case IMAGE_BOOT_CODE:
3999 img_optype = OPTYPE_REDBOOT;
4001 case IMAGE_OPTION_ROM_ISCSI:
4002 img_optype = OPTYPE_BIOS;
4004 case IMAGE_OPTION_ROM_PXE:
4005 img_optype = OPTYPE_PXE_BIOS;
4007 case IMAGE_OPTION_ROM_FCoE:
4008 img_optype = OPTYPE_FCOE_BIOS;
4010 case IMAGE_FIRMWARE_BACKUP_iSCSI:
4011 img_optype = OPTYPE_ISCSI_BACKUP;
4014 img_optype = OPTYPE_NCSI_FW;
4016 case IMAGE_FLASHISM_JUMPVECTOR:
4017 img_optype = OPTYPE_FLASHISM_JUMPVECTOR;
4019 case IMAGE_FIRMWARE_PHY:
4020 img_optype = OPTYPE_SH_PHY_FW;
4022 case IMAGE_REDBOOT_DIR:
4023 img_optype = OPTYPE_REDBOOT_DIR;
4025 case IMAGE_REDBOOT_CONFIG:
4026 img_optype = OPTYPE_REDBOOT_CONFIG;
4029 img_optype = OPTYPE_UFI_DIR;
4038 static int be_flash_skyhawk(struct be_adapter *adapter,
4039 const struct firmware *fw,
4040 struct be_dma_mem *flash_cmd, int num_of_images)
4042 int img_hdrs_size = num_of_images * sizeof(struct image_hdr);
4043 struct device *dev = &adapter->pdev->dev;
4044 struct flash_section_info *fsec = NULL;
4045 u32 img_offset, img_size, img_type;
4046 int status, i, filehdr_size;
4047 bool crc_match, old_fw_img;
4051 filehdr_size = sizeof(struct flash_file_hdr_g3);
4052 fsec = get_fsec_info(adapter, filehdr_size + img_hdrs_size, fw);
4054 dev_err(dev, "Invalid Cookie. FW image may be corrupted\n");
4058 for (i = 0; i < le32_to_cpu(fsec->fsec_hdr.num_images); i++) {
4059 img_offset = le32_to_cpu(fsec->fsec_entry[i].offset);
4060 img_size = le32_to_cpu(fsec->fsec_entry[i].pad_size);
4061 img_type = le32_to_cpu(fsec->fsec_entry[i].type);
4062 img_optype = be_get_img_optype(fsec->fsec_entry[i]);
4063 old_fw_img = fsec->fsec_entry[i].optype == 0xFFFF;
4065 if (img_optype == 0xFFFF)
4067 /* Don't bother verifying CRC if an old FW image is being
4073 status = be_check_flash_crc(adapter, fw->data, img_offset,
4074 img_size, filehdr_size +
4075 img_hdrs_size, img_optype,
4077 /* The current FW image on the card does not recognize the new
4078 * FLASH op_type. The FW download is partially complete.
4079 * Reboot the server now to enable FW image to recognize the
4080 * new FLASH op_type. To complete the remaining process,
4081 * download the same FW again after the reboot.
4083 if (base_status(status) == MCC_STATUS_ILLEGAL_REQUEST ||
4084 base_status(status) == MCC_STATUS_ILLEGAL_FIELD) {
4085 dev_err(dev, "Flash incomplete. Reset the server\n");
4086 dev_err(dev, "Download FW image again after reset\n");
4088 } else if (status) {
4089 dev_err(dev, "Could not get CRC for 0x%x region\n",
4098 p = fw->data + filehdr_size + img_offset + img_hdrs_size;
4099 if (p + img_size > fw->data + fw->size)
4102 status = be_flash(adapter, p, flash_cmd, img_optype, img_size);
4103 /* For old FW images ignore ILLEGAL_FIELD error or errors on
4107 (base_status(status) == MCC_STATUS_ILLEGAL_FIELD ||
4108 (img_optype == OPTYPE_UFI_DIR &&
4109 base_status(status) == MCC_STATUS_FAILED))) {
4111 } else if (status) {
4112 dev_err(dev, "Flashing section type 0x%x failed\n",
4120 static int lancer_fw_download(struct be_adapter *adapter,
4121 const struct firmware *fw)
4123 #define LANCER_FW_DOWNLOAD_CHUNK (32 * 1024)
4124 #define LANCER_FW_DOWNLOAD_LOCATION "/prg"
4125 struct device *dev = &adapter->pdev->dev;
4126 struct be_dma_mem flash_cmd;
4127 const u8 *data_ptr = NULL;
4128 u8 *dest_image_ptr = NULL;
4129 size_t image_size = 0;
4131 u32 data_written = 0;
4137 if (!IS_ALIGNED(fw->size, sizeof(u32))) {
4138 dev_err(dev, "FW image size should be multiple of 4\n");
4142 flash_cmd.size = sizeof(struct lancer_cmd_req_write_object)
4143 + LANCER_FW_DOWNLOAD_CHUNK;
4144 flash_cmd.va = dma_alloc_coherent(dev, flash_cmd.size,
4145 &flash_cmd.dma, GFP_KERNEL);
4149 dest_image_ptr = flash_cmd.va +
4150 sizeof(struct lancer_cmd_req_write_object);
4151 image_size = fw->size;
4152 data_ptr = fw->data;
4154 while (image_size) {
4155 chunk_size = min_t(u32, image_size, LANCER_FW_DOWNLOAD_CHUNK);
4157 /* Copy the image chunk content. */
4158 memcpy(dest_image_ptr, data_ptr, chunk_size);
4160 status = lancer_cmd_write_object(adapter, &flash_cmd,
4162 LANCER_FW_DOWNLOAD_LOCATION,
4163 &data_written, &change_status,
4168 offset += data_written;
4169 data_ptr += data_written;
4170 image_size -= data_written;
4174 /* Commit the FW written */
4175 status = lancer_cmd_write_object(adapter, &flash_cmd,
4177 LANCER_FW_DOWNLOAD_LOCATION,
4178 &data_written, &change_status,
4182 dma_free_coherent(dev, flash_cmd.size, flash_cmd.va, flash_cmd.dma);
4184 dev_err(dev, "Firmware load error\n");
4185 return be_cmd_status(status);
4188 dev_info(dev, "Firmware flashed successfully\n");
4190 if (change_status == LANCER_FW_RESET_NEEDED) {
4191 dev_info(dev, "Resetting adapter to activate new FW\n");
4192 status = lancer_physdev_ctrl(adapter,
4193 PHYSDEV_CONTROL_FW_RESET_MASK);
4195 dev_err(dev, "Adapter busy, could not reset FW\n");
4196 dev_err(dev, "Reboot server to activate new FW\n");
4198 } else if (change_status != LANCER_NO_RESET_NEEDED) {
4199 dev_info(dev, "Reboot server to activate new FW\n");
4207 #define UFI_TYPE3R 10
4209 static int be_get_ufi_type(struct be_adapter *adapter,
4210 struct flash_file_hdr_g3 *fhdr)
4213 goto be_get_ufi_exit;
4215 if (skyhawk_chip(adapter) && fhdr->build[0] == '4')
4217 else if (BE3_chip(adapter) && fhdr->build[0] == '3') {
4218 if (fhdr->asic_type_rev == 0x10)
4222 } else if (BE2_chip(adapter) && fhdr->build[0] == '2')
4226 dev_err(&adapter->pdev->dev,
4227 "UFI and Interface are not compatible for flashing\n");
4231 static int be_fw_download(struct be_adapter *adapter, const struct firmware* fw)
4233 struct flash_file_hdr_g3 *fhdr3;
4234 struct image_hdr *img_hdr_ptr = NULL;
4235 struct be_dma_mem flash_cmd;
4237 int status = 0, i = 0, num_imgs = 0, ufi_type = 0;
4239 flash_cmd.size = sizeof(struct be_cmd_write_flashrom);
4240 flash_cmd.va = dma_alloc_coherent(&adapter->pdev->dev, flash_cmd.size,
4241 &flash_cmd.dma, GFP_KERNEL);
4242 if (!flash_cmd.va) {
4248 fhdr3 = (struct flash_file_hdr_g3 *)p;
4250 ufi_type = be_get_ufi_type(adapter, fhdr3);
4252 num_imgs = le32_to_cpu(fhdr3->num_imgs);
4253 for (i = 0; i < num_imgs; i++) {
4254 img_hdr_ptr = (struct image_hdr *)(fw->data +
4255 (sizeof(struct flash_file_hdr_g3) +
4256 i * sizeof(struct image_hdr)));
4257 if (le32_to_cpu(img_hdr_ptr->imageid) == 1) {
4260 status = be_flash_skyhawk(adapter, fw,
4261 &flash_cmd, num_imgs);
4264 status = be_flash_BEx(adapter, fw, &flash_cmd,
4268 /* Do not flash this ufi on BE3-R cards */
4269 if (adapter->asic_rev < 0x10)
4270 status = be_flash_BEx(adapter, fw,
4275 dev_err(&adapter->pdev->dev,
4276 "Can't load BE3 UFI on BE3R\n");
4282 if (ufi_type == UFI_TYPE2)
4283 status = be_flash_BEx(adapter, fw, &flash_cmd, 0);
4284 else if (ufi_type == -1)
4287 dma_free_coherent(&adapter->pdev->dev, flash_cmd.size, flash_cmd.va,
4290 dev_err(&adapter->pdev->dev, "Firmware load error\n");
4294 dev_info(&adapter->pdev->dev, "Firmware flashed successfully\n");
4300 int be_load_fw(struct be_adapter *adapter, u8 *fw_file)
4302 const struct firmware *fw;
4305 if (!netif_running(adapter->netdev)) {
4306 dev_err(&adapter->pdev->dev,
4307 "Firmware load not allowed (interface is down)\n");
4311 status = request_firmware(&fw, fw_file, &adapter->pdev->dev);
4315 dev_info(&adapter->pdev->dev, "Flashing firmware file %s\n", fw_file);
4317 if (lancer_chip(adapter))
4318 status = lancer_fw_download(adapter, fw);
4320 status = be_fw_download(adapter, fw);
4323 be_cmd_get_fw_ver(adapter);
4326 release_firmware(fw);
4330 static int be_ndo_bridge_setlink(struct net_device *dev, struct nlmsghdr *nlh,
4333 struct be_adapter *adapter = netdev_priv(dev);
4334 struct nlattr *attr, *br_spec;
4339 if (!sriov_enabled(adapter))
4342 br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
4346 nla_for_each_nested(attr, br_spec, rem) {
4347 if (nla_type(attr) != IFLA_BRIDGE_MODE)
4350 if (nla_len(attr) < sizeof(mode))
4353 mode = nla_get_u16(attr);
4354 if (mode != BRIDGE_MODE_VEPA && mode != BRIDGE_MODE_VEB)
4357 status = be_cmd_set_hsw_config(adapter, 0, 0,
4359 mode == BRIDGE_MODE_VEPA ?
4360 PORT_FWD_TYPE_VEPA :
4365 dev_info(&adapter->pdev->dev, "enabled switch mode: %s\n",
4366 mode == BRIDGE_MODE_VEPA ? "VEPA" : "VEB");
4371 dev_err(&adapter->pdev->dev, "Failed to set switch mode %s\n",
4372 mode == BRIDGE_MODE_VEPA ? "VEPA" : "VEB");
4377 static int be_ndo_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
4378 struct net_device *dev, u32 filter_mask)
4380 struct be_adapter *adapter = netdev_priv(dev);
4384 if (!sriov_enabled(adapter))
4387 /* BE and Lancer chips support VEB mode only */
4388 if (BEx_chip(adapter) || lancer_chip(adapter)) {
4389 hsw_mode = PORT_FWD_TYPE_VEB;
4391 status = be_cmd_get_hsw_config(adapter, NULL, 0,
4392 adapter->if_handle, &hsw_mode);
4397 return ndo_dflt_bridge_getlink(skb, pid, seq, dev,
4398 hsw_mode == PORT_FWD_TYPE_VEPA ?
4399 BRIDGE_MODE_VEPA : BRIDGE_MODE_VEB,
4403 #ifdef CONFIG_BE2NET_VXLAN
4404 /* VxLAN offload Notes:
4406 * The stack defines tunnel offload flags (hw_enc_features) for IP and doesn't
4407 * distinguish various types of transports (VxLAN, GRE, NVGRE ..). So, offload
4408 * is expected to work across all types of IP tunnels once exported. Skyhawk
4409 * supports offloads for either VxLAN or NVGRE, exclusively. So we export VxLAN
4410 * offloads in hw_enc_features only when a VxLAN port is added. If other (non
4411 * VxLAN) tunnels are configured while VxLAN offloads are enabled, offloads for
4412 * those other tunnels are unexported on the fly through ndo_features_check().
4414 * Skyhawk supports VxLAN offloads only for one UDP dport. So, if the stack
4415 * adds more than one port, disable offloads and don't re-enable them again
4416 * until after all the tunnels are removed.
4418 static void be_add_vxlan_port(struct net_device *netdev, sa_family_t sa_family,
4421 struct be_adapter *adapter = netdev_priv(netdev);
4422 struct device *dev = &adapter->pdev->dev;
4425 if (lancer_chip(adapter) || BEx_chip(adapter))
4428 if (adapter->flags & BE_FLAGS_VXLAN_OFFLOADS) {
4430 "Only one UDP port supported for VxLAN offloads\n");
4431 dev_info(dev, "Disabling VxLAN offloads\n");
4432 adapter->vxlan_port_count++;
4436 if (adapter->vxlan_port_count++ >= 1)
4439 status = be_cmd_manage_iface(adapter, adapter->if_handle,
4440 OP_CONVERT_NORMAL_TO_TUNNEL);
4442 dev_warn(dev, "Failed to convert normal interface to tunnel\n");
4446 status = be_cmd_set_vxlan_port(adapter, port);
4448 dev_warn(dev, "Failed to add VxLAN port\n");
4451 adapter->flags |= BE_FLAGS_VXLAN_OFFLOADS;
4452 adapter->vxlan_port = port;
4454 netdev->hw_enc_features |= NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM |
4455 NETIF_F_TSO | NETIF_F_TSO6 |
4456 NETIF_F_GSO_UDP_TUNNEL;
4457 netdev->hw_features |= NETIF_F_GSO_UDP_TUNNEL;
4458 netdev->features |= NETIF_F_GSO_UDP_TUNNEL;
4460 dev_info(dev, "Enabled VxLAN offloads for UDP port %d\n",
4464 be_disable_vxlan_offloads(adapter);
4467 static void be_del_vxlan_port(struct net_device *netdev, sa_family_t sa_family,
4470 struct be_adapter *adapter = netdev_priv(netdev);
4472 if (lancer_chip(adapter) || BEx_chip(adapter))
4475 if (adapter->vxlan_port != port)
4478 be_disable_vxlan_offloads(adapter);
4480 dev_info(&adapter->pdev->dev,
4481 "Disabled VxLAN offloads for UDP port %d\n",
4484 adapter->vxlan_port_count--;
4487 static netdev_features_t be_features_check(struct sk_buff *skb,
4488 struct net_device *dev,
4489 netdev_features_t features)
4491 struct be_adapter *adapter = netdev_priv(dev);
4494 /* The code below restricts offload features for some tunneled packets.
4495 * Offload features for normal (non tunnel) packets are unchanged.
4497 if (!skb->encapsulation ||
4498 !(adapter->flags & BE_FLAGS_VXLAN_OFFLOADS))
4501 /* It's an encapsulated packet and VxLAN offloads are enabled. We
4502 * should disable tunnel offload features if it's not a VxLAN packet,
4503 * as tunnel offloads have been enabled only for VxLAN. This is done to
4504 * allow other tunneled traffic like GRE work fine while VxLAN
4505 * offloads are configured in Skyhawk-R.
4507 switch (vlan_get_protocol(skb)) {
4508 case htons(ETH_P_IP):
4509 l4_hdr = ip_hdr(skb)->protocol;
4511 case htons(ETH_P_IPV6):
4512 l4_hdr = ipv6_hdr(skb)->nexthdr;
4518 if (l4_hdr != IPPROTO_UDP ||
4519 skb->inner_protocol_type != ENCAP_TYPE_ETHER ||
4520 skb->inner_protocol != htons(ETH_P_TEB) ||
4521 skb_inner_mac_header(skb) - skb_transport_header(skb) !=
4522 sizeof(struct udphdr) + sizeof(struct vxlanhdr))
4523 return features & ~(NETIF_F_ALL_CSUM | NETIF_F_GSO_MASK);
4529 static const struct net_device_ops be_netdev_ops = {
4530 .ndo_open = be_open,
4531 .ndo_stop = be_close,
4532 .ndo_start_xmit = be_xmit,
4533 .ndo_set_rx_mode = be_set_rx_mode,
4534 .ndo_set_mac_address = be_mac_addr_set,
4535 .ndo_change_mtu = be_change_mtu,
4536 .ndo_get_stats64 = be_get_stats64,
4537 .ndo_validate_addr = eth_validate_addr,
4538 .ndo_vlan_rx_add_vid = be_vlan_add_vid,
4539 .ndo_vlan_rx_kill_vid = be_vlan_rem_vid,
4540 .ndo_set_vf_mac = be_set_vf_mac,
4541 .ndo_set_vf_vlan = be_set_vf_vlan,
4542 .ndo_set_vf_rate = be_set_vf_tx_rate,
4543 .ndo_get_vf_config = be_get_vf_config,
4544 .ndo_set_vf_link_state = be_set_vf_link_state,
4545 #ifdef CONFIG_NET_POLL_CONTROLLER
4546 .ndo_poll_controller = be_netpoll,
4548 .ndo_bridge_setlink = be_ndo_bridge_setlink,
4549 .ndo_bridge_getlink = be_ndo_bridge_getlink,
4550 #ifdef CONFIG_NET_RX_BUSY_POLL
4551 .ndo_busy_poll = be_busy_poll,
4553 #ifdef CONFIG_BE2NET_VXLAN
4554 .ndo_add_vxlan_port = be_add_vxlan_port,
4555 .ndo_del_vxlan_port = be_del_vxlan_port,
4556 .ndo_features_check = be_features_check,
4560 static void be_netdev_init(struct net_device *netdev)
4562 struct be_adapter *adapter = netdev_priv(netdev);
4564 netdev->hw_features |= NETIF_F_SG | NETIF_F_TSO | NETIF_F_TSO6 |
4565 NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM | NETIF_F_RXCSUM |
4566 NETIF_F_HW_VLAN_CTAG_TX;
4567 if (be_multi_rxq(adapter))
4568 netdev->hw_features |= NETIF_F_RXHASH;
4570 netdev->features |= netdev->hw_features |
4571 NETIF_F_HW_VLAN_CTAG_RX | NETIF_F_HW_VLAN_CTAG_FILTER;
4573 netdev->vlan_features |= NETIF_F_SG | NETIF_F_TSO | NETIF_F_TSO6 |
4574 NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM;
4576 netdev->priv_flags |= IFF_UNICAST_FLT;
4578 netdev->flags |= IFF_MULTICAST;
4580 netif_set_gso_max_size(netdev, 65535 - ETH_HLEN);
4582 netdev->netdev_ops = &be_netdev_ops;
4584 netdev->ethtool_ops = &be_ethtool_ops;
4587 static void be_unmap_pci_bars(struct be_adapter *adapter)
4590 pci_iounmap(adapter->pdev, adapter->csr);
4592 pci_iounmap(adapter->pdev, adapter->db);
4595 static int db_bar(struct be_adapter *adapter)
4597 if (lancer_chip(adapter) || !be_physfn(adapter))
4603 static int be_roce_map_pci_bars(struct be_adapter *adapter)
4605 if (skyhawk_chip(adapter)) {
4606 adapter->roce_db.size = 4096;
4607 adapter->roce_db.io_addr = pci_resource_start(adapter->pdev,
4609 adapter->roce_db.total_size = pci_resource_len(adapter->pdev,
4615 static int be_map_pci_bars(struct be_adapter *adapter)
4619 if (BEx_chip(adapter) && be_physfn(adapter)) {
4620 adapter->csr = pci_iomap(adapter->pdev, 2, 0);
4625 addr = pci_iomap(adapter->pdev, db_bar(adapter), 0);
4630 be_roce_map_pci_bars(adapter);
4634 dev_err(&adapter->pdev->dev, "Error in mapping PCI BARs\n");
4635 be_unmap_pci_bars(adapter);
4639 static void be_ctrl_cleanup(struct be_adapter *adapter)
4641 struct be_dma_mem *mem = &adapter->mbox_mem_alloced;
4643 be_unmap_pci_bars(adapter);
4646 dma_free_coherent(&adapter->pdev->dev, mem->size, mem->va,
4649 mem = &adapter->rx_filter;
4651 dma_free_coherent(&adapter->pdev->dev, mem->size, mem->va,
4655 static int be_ctrl_init(struct be_adapter *adapter)
4657 struct be_dma_mem *mbox_mem_alloc = &adapter->mbox_mem_alloced;
4658 struct be_dma_mem *mbox_mem_align = &adapter->mbox_mem;
4659 struct be_dma_mem *rx_filter = &adapter->rx_filter;
4663 pci_read_config_dword(adapter->pdev, SLI_INTF_REG_OFFSET, &sli_intf);
4664 adapter->sli_family = (sli_intf & SLI_INTF_FAMILY_MASK) >>
4665 SLI_INTF_FAMILY_SHIFT;
4666 adapter->virtfn = (sli_intf & SLI_INTF_FT_MASK) ? 1 : 0;
4668 status = be_map_pci_bars(adapter);
4672 mbox_mem_alloc->size = sizeof(struct be_mcc_mailbox) + 16;
4673 mbox_mem_alloc->va = dma_alloc_coherent(&adapter->pdev->dev,
4674 mbox_mem_alloc->size,
4675 &mbox_mem_alloc->dma,
4677 if (!mbox_mem_alloc->va) {
4679 goto unmap_pci_bars;
4681 mbox_mem_align->size = sizeof(struct be_mcc_mailbox);
4682 mbox_mem_align->va = PTR_ALIGN(mbox_mem_alloc->va, 16);
4683 mbox_mem_align->dma = PTR_ALIGN(mbox_mem_alloc->dma, 16);
4684 memset(mbox_mem_align->va, 0, sizeof(struct be_mcc_mailbox));
4686 rx_filter->size = sizeof(struct be_cmd_req_rx_filter);
4687 rx_filter->va = dma_zalloc_coherent(&adapter->pdev->dev,
4688 rx_filter->size, &rx_filter->dma,
4690 if (!rx_filter->va) {
4695 mutex_init(&adapter->mbox_lock);
4696 spin_lock_init(&adapter->mcc_lock);
4697 spin_lock_init(&adapter->mcc_cq_lock);
4699 init_completion(&adapter->et_cmd_compl);
4700 pci_save_state(adapter->pdev);
4704 dma_free_coherent(&adapter->pdev->dev, mbox_mem_alloc->size,
4705 mbox_mem_alloc->va, mbox_mem_alloc->dma);
4708 be_unmap_pci_bars(adapter);
4714 static void be_stats_cleanup(struct be_adapter *adapter)
4716 struct be_dma_mem *cmd = &adapter->stats_cmd;
4719 dma_free_coherent(&adapter->pdev->dev, cmd->size,
4723 static int be_stats_init(struct be_adapter *adapter)
4725 struct be_dma_mem *cmd = &adapter->stats_cmd;
4727 if (lancer_chip(adapter))
4728 cmd->size = sizeof(struct lancer_cmd_req_pport_stats);
4729 else if (BE2_chip(adapter))
4730 cmd->size = sizeof(struct be_cmd_req_get_stats_v0);
4731 else if (BE3_chip(adapter))
4732 cmd->size = sizeof(struct be_cmd_req_get_stats_v1);
4734 /* ALL non-BE ASICs */
4735 cmd->size = sizeof(struct be_cmd_req_get_stats_v2);
4737 cmd->va = dma_zalloc_coherent(&adapter->pdev->dev, cmd->size, &cmd->dma,
4744 static void be_remove(struct pci_dev *pdev)
4746 struct be_adapter *adapter = pci_get_drvdata(pdev);
4751 be_roce_dev_remove(adapter);
4752 be_intr_set(adapter, false);
4754 cancel_delayed_work_sync(&adapter->func_recovery_work);
4756 unregister_netdev(adapter->netdev);
4760 /* tell fw we're done with firing cmds */
4761 be_cmd_fw_clean(adapter);
4763 be_stats_cleanup(adapter);
4765 be_ctrl_cleanup(adapter);
4767 pci_disable_pcie_error_reporting(pdev);
4769 pci_release_regions(pdev);
4770 pci_disable_device(pdev);
4772 free_netdev(adapter->netdev);
4775 static int be_get_initial_config(struct be_adapter *adapter)
4779 status = be_cmd_get_cntl_attributes(adapter);
4783 /* Must be a power of 2 or else MODULO will BUG_ON */
4784 adapter->be_get_temp_freq = 64;
4786 if (BEx_chip(adapter)) {
4787 level = be_cmd_get_fw_log_level(adapter);
4788 adapter->msg_enable =
4789 level <= FW_LOG_LEVEL_DEFAULT ? NETIF_MSG_HW : 0;
4792 adapter->cfg_num_qs = netif_get_num_default_rss_queues();
4796 static int lancer_recover_func(struct be_adapter *adapter)
4798 struct device *dev = &adapter->pdev->dev;
4801 status = lancer_test_and_set_rdy_state(adapter);
4805 if (netif_running(adapter->netdev))
4806 be_close(adapter->netdev);
4810 be_clear_all_error(adapter);
4812 status = be_setup(adapter);
4816 if (netif_running(adapter->netdev)) {
4817 status = be_open(adapter->netdev);
4822 dev_err(dev, "Adapter recovery successful\n");
4825 if (status == -EAGAIN)
4826 dev_err(dev, "Waiting for resource provisioning\n");
4828 dev_err(dev, "Adapter recovery failed\n");
4833 static void be_func_recovery_task(struct work_struct *work)
4835 struct be_adapter *adapter =
4836 container_of(work, struct be_adapter, func_recovery_work.work);
4839 be_detect_error(adapter);
4841 if (adapter->hw_error && lancer_chip(adapter)) {
4843 netif_device_detach(adapter->netdev);
4846 status = lancer_recover_func(adapter);
4848 netif_device_attach(adapter->netdev);
4851 /* In Lancer, for all errors other than provisioning error (-EAGAIN),
4852 * no need to attempt further recovery.
4854 if (!status || status == -EAGAIN)
4855 schedule_delayed_work(&adapter->func_recovery_work,
4856 msecs_to_jiffies(1000));
4859 static void be_worker(struct work_struct *work)
4861 struct be_adapter *adapter =
4862 container_of(work, struct be_adapter, work.work);
4863 struct be_rx_obj *rxo;
4866 /* when interrupts are not yet enabled, just reap any pending
4867 * mcc completions */
4868 if (!netif_running(adapter->netdev)) {
4870 be_process_mcc(adapter);
4875 if (!adapter->stats_cmd_sent) {
4876 if (lancer_chip(adapter))
4877 lancer_cmd_get_pport_stats(adapter,
4878 &adapter->stats_cmd);
4880 be_cmd_get_stats(adapter, &adapter->stats_cmd);
4883 if (be_physfn(adapter) &&
4884 MODULO(adapter->work_counter, adapter->be_get_temp_freq) == 0)
4885 be_cmd_get_die_temperature(adapter);
4887 for_all_rx_queues(adapter, rxo, i) {
4888 /* Replenish RX-queues starved due to memory
4889 * allocation failures.
4891 if (rxo->rx_post_starved)
4892 be_post_rx_frags(rxo, GFP_KERNEL, MAX_RX_POST);
4895 be_eqd_update(adapter);
4898 adapter->work_counter++;
4899 schedule_delayed_work(&adapter->work, msecs_to_jiffies(1000));
4902 /* If any VFs are already enabled don't FLR the PF */
4903 static bool be_reset_required(struct be_adapter *adapter)
4905 return pci_num_vf(adapter->pdev) ? false : true;
4908 static char *mc_name(struct be_adapter *adapter)
4910 char *str = ""; /* default */
4912 switch (adapter->mc_type) {
4938 static inline char *func_name(struct be_adapter *adapter)
4940 return be_physfn(adapter) ? "PF" : "VF";
4943 static int be_probe(struct pci_dev *pdev, const struct pci_device_id *pdev_id)
4946 struct be_adapter *adapter;
4947 struct net_device *netdev;
4950 dev_info(&pdev->dev, "%s version is %s\n", DRV_NAME, DRV_VER);
4952 status = pci_enable_device(pdev);
4956 status = pci_request_regions(pdev, DRV_NAME);
4959 pci_set_master(pdev);
4961 netdev = alloc_etherdev_mqs(sizeof(*adapter), MAX_TX_QS, MAX_RX_QS);
4966 adapter = netdev_priv(netdev);
4967 adapter->pdev = pdev;
4968 pci_set_drvdata(pdev, adapter);
4969 adapter->netdev = netdev;
4970 SET_NETDEV_DEV(netdev, &pdev->dev);
4972 status = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(64));
4974 netdev->features |= NETIF_F_HIGHDMA;
4976 status = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(32));
4978 dev_err(&pdev->dev, "Could not set PCI DMA Mask\n");
4983 status = pci_enable_pcie_error_reporting(pdev);
4985 dev_info(&pdev->dev, "PCIe error reporting enabled\n");
4987 status = be_ctrl_init(adapter);
4991 /* sync up with fw's ready state */
4992 if (be_physfn(adapter)) {
4993 status = be_fw_wait_ready(adapter);
4998 if (be_reset_required(adapter)) {
4999 status = be_cmd_reset_function(adapter);
5003 /* Wait for interrupts to quiesce after an FLR */
5007 /* Allow interrupts for other ULPs running on NIC function */
5008 be_intr_set(adapter, true);
5010 /* tell fw we're ready to fire cmds */
5011 status = be_cmd_fw_init(adapter);
5015 status = be_stats_init(adapter);
5019 status = be_get_initial_config(adapter);
5023 INIT_DELAYED_WORK(&adapter->work, be_worker);
5024 INIT_DELAYED_WORK(&adapter->func_recovery_work, be_func_recovery_task);
5025 adapter->rx_fc = true;
5026 adapter->tx_fc = true;
5028 status = be_setup(adapter);
5032 be_netdev_init(netdev);
5033 status = register_netdev(netdev);
5037 be_roce_dev_add(adapter);
5039 schedule_delayed_work(&adapter->func_recovery_work,
5040 msecs_to_jiffies(1000));
5042 be_cmd_query_port_name(adapter, &port_name);
5044 dev_info(&pdev->dev, "%s: %s %s port %c\n", nic_name(pdev),
5045 func_name(adapter), mc_name(adapter), port_name);
5052 be_stats_cleanup(adapter);
5054 be_ctrl_cleanup(adapter);
5056 free_netdev(netdev);
5058 pci_release_regions(pdev);
5060 pci_disable_device(pdev);
5062 dev_err(&pdev->dev, "%s initialization failed\n", nic_name(pdev));
5066 static int be_suspend(struct pci_dev *pdev, pm_message_t state)
5068 struct be_adapter *adapter = pci_get_drvdata(pdev);
5069 struct net_device *netdev = adapter->netdev;
5071 if (adapter->wol_en)
5072 be_setup_wol(adapter, true);
5074 be_intr_set(adapter, false);
5075 cancel_delayed_work_sync(&adapter->func_recovery_work);
5077 netif_device_detach(netdev);
5078 if (netif_running(netdev)) {
5085 pci_save_state(pdev);
5086 pci_disable_device(pdev);
5087 pci_set_power_state(pdev, pci_choose_state(pdev, state));
5091 static int be_resume(struct pci_dev *pdev)
5094 struct be_adapter *adapter = pci_get_drvdata(pdev);
5095 struct net_device *netdev = adapter->netdev;
5097 netif_device_detach(netdev);
5099 status = pci_enable_device(pdev);
5103 pci_set_power_state(pdev, PCI_D0);
5104 pci_restore_state(pdev);
5106 status = be_fw_wait_ready(adapter);
5110 status = be_cmd_reset_function(adapter);
5114 be_intr_set(adapter, true);
5115 /* tell fw we're ready to fire cmds */
5116 status = be_cmd_fw_init(adapter);
5121 if (netif_running(netdev)) {
5127 schedule_delayed_work(&adapter->func_recovery_work,
5128 msecs_to_jiffies(1000));
5129 netif_device_attach(netdev);
5131 if (adapter->wol_en)
5132 be_setup_wol(adapter, false);
5138 * An FLR will stop BE from DMAing any data.
5140 static void be_shutdown(struct pci_dev *pdev)
5142 struct be_adapter *adapter = pci_get_drvdata(pdev);
5147 be_roce_dev_shutdown(adapter);
5148 cancel_delayed_work_sync(&adapter->work);
5149 cancel_delayed_work_sync(&adapter->func_recovery_work);
5151 netif_device_detach(adapter->netdev);
5153 be_cmd_reset_function(adapter);
5155 pci_disable_device(pdev);
5158 static pci_ers_result_t be_eeh_err_detected(struct pci_dev *pdev,
5159 pci_channel_state_t state)
5161 struct be_adapter *adapter = pci_get_drvdata(pdev);
5162 struct net_device *netdev = adapter->netdev;
5164 dev_err(&adapter->pdev->dev, "EEH error detected\n");
5166 if (!adapter->eeh_error) {
5167 adapter->eeh_error = true;
5169 cancel_delayed_work_sync(&adapter->func_recovery_work);
5172 netif_device_detach(netdev);
5173 if (netif_running(netdev))
5180 if (state == pci_channel_io_perm_failure)
5181 return PCI_ERS_RESULT_DISCONNECT;
5183 pci_disable_device(pdev);
5185 /* The error could cause the FW to trigger a flash debug dump.
5186 * Resetting the card while flash dump is in progress
5187 * can cause it not to recover; wait for it to finish.
5188 * Wait only for first function as it is needed only once per
5191 if (pdev->devfn == 0)
5194 return PCI_ERS_RESULT_NEED_RESET;
5197 static pci_ers_result_t be_eeh_reset(struct pci_dev *pdev)
5199 struct be_adapter *adapter = pci_get_drvdata(pdev);
5202 dev_info(&adapter->pdev->dev, "EEH reset\n");
5204 status = pci_enable_device(pdev);
5206 return PCI_ERS_RESULT_DISCONNECT;
5208 pci_set_master(pdev);
5209 pci_set_power_state(pdev, PCI_D0);
5210 pci_restore_state(pdev);
5212 /* Check if card is ok and fw is ready */
5213 dev_info(&adapter->pdev->dev,
5214 "Waiting for FW to be ready after EEH reset\n");
5215 status = be_fw_wait_ready(adapter);
5217 return PCI_ERS_RESULT_DISCONNECT;
5219 pci_cleanup_aer_uncorrect_error_status(pdev);
5220 be_clear_all_error(adapter);
5221 return PCI_ERS_RESULT_RECOVERED;
5224 static void be_eeh_resume(struct pci_dev *pdev)
5227 struct be_adapter *adapter = pci_get_drvdata(pdev);
5228 struct net_device *netdev = adapter->netdev;
5230 dev_info(&adapter->pdev->dev, "EEH resume\n");
5232 pci_save_state(pdev);
5234 status = be_cmd_reset_function(adapter);
5238 /* On some BE3 FW versions, after a HW reset,
5239 * interrupts will remain disabled for each function.
5240 * So, explicitly enable interrupts
5242 be_intr_set(adapter, true);
5244 /* tell fw we're ready to fire cmds */
5245 status = be_cmd_fw_init(adapter);
5249 status = be_setup(adapter);
5253 if (netif_running(netdev)) {
5254 status = be_open(netdev);
5259 schedule_delayed_work(&adapter->func_recovery_work,
5260 msecs_to_jiffies(1000));
5261 netif_device_attach(netdev);
5264 dev_err(&adapter->pdev->dev, "EEH resume failed\n");
5267 static const struct pci_error_handlers be_eeh_handlers = {
5268 .error_detected = be_eeh_err_detected,
5269 .slot_reset = be_eeh_reset,
5270 .resume = be_eeh_resume,
5273 static struct pci_driver be_driver = {
5275 .id_table = be_dev_ids,
5277 .remove = be_remove,
5278 .suspend = be_suspend,
5279 .resume = be_resume,
5280 .shutdown = be_shutdown,
5281 .err_handler = &be_eeh_handlers
5284 static int __init be_init_module(void)
5286 if (rx_frag_size != 8192 && rx_frag_size != 4096 &&
5287 rx_frag_size != 2048) {
5288 printk(KERN_WARNING DRV_NAME
5289 " : Module param rx_frag_size must be 2048/4096/8192."
5291 rx_frag_size = 2048;
5294 return pci_register_driver(&be_driver);
5296 module_init(be_init_module);
5298 static void __exit be_exit_module(void)
5300 pci_unregister_driver(&be_driver);
5302 module_exit(be_exit_module);